ArticlePDF AvailableLiterature Review

Friston, K.J.: The free-energy principle: a unified brain theory? Nat. Rev. Neurosci. 11, 127-138

Authors:

Abstract

A free-energy principle has been proposed recently that accounts for action, perception and learning. This Review looks at some key brain theories in the biological (for example, neural Darwinism) and physical (for example, information theory and optimal control theory) sciences from the free-energy perspective. Crucially, one key theme runs through each of these theories - optimization. Furthermore, if we look closely at what is optimized, the same quantity keeps emerging, namely value (expected reward, expected utility) or its complement, surprise (prediction error, expected cost). This is the quantity that is optimized under the free-energy principle, which suggests that several global brain theories might be unified within a free-energy framework.
Despite the wealth of empirical data in neuroscience,
there are relatively few global theories about how the
brain works. A recently proposed free-energy principle
for adaptive systems tries to provide a unified account
of action, perception and learning. Although this prin-
ciple has been portrayed as a unified brain theory1, its
capacity to unify different perspectives on brain function
has yet to be established. This Review attempts to place
some key theories within the free-energy framework, in
the hope of identifying common themes. I first review
the free-energy principle and then deconstruct several
global brain theories to show how they all speak to the
same underlying idea.
The free-energy principle
The free-energy principle (BOX 1) says that any self-
organizing system that is at equilibrium with its environ-
ment must minimize its free energy2. The principle is
essentially a mathematical formulation of how adaptive
systems (that is, biological agents, like animals or brains)
resist a natural tendency to disorder3–6. What follows is
a non-mathematical treatment of the motivation and
implications of the principle. We will see that although the
motivation is quite straightforward, the implications are
complicated and diverse. This diversity allows the prin-
ciple to account for many aspects of brain structure and
function and lends it the potential to unify different per-
spectives on how the brain works. In subsequent sections,
I discuss how the principle can be applied to neuronal
systems as viewed from these perspectives. This Review
starts in a rather abstract and technical way but then tries
to unpack the basic idea in more familiar terms.
Motivation: resisting a tendency to disorder. The
defining characteristic of biological systems is that
they maintain their states and form in the face of a
constantly changing environment3–6. From the point
of view of the brain, the environment includes both
the external and the internal milieu. This maintenance
of order is seen at many levels and distinguishes bio-
logical from other self-organizing systems; indeed, the
physiology of biological systems can be reduced almost
entirely to their homeostasis7. More precisely, the rep-
ertoire of physiological and sensory states in which an
organism can be is limited, and these states define the
organisms phenotype. Mathematically, this means that
the probability of these (interoceptive and exterocep-
tive) sensory states must have low entropy; in other
words, there is a high probability that a system will
be in any of a small number of states, and a low prob-
ability that it will be in the remaining states. Entropy
is also the average self information or surprise8
(more formally, it is the negative log-probability of an
outcome). Here, ‘a fish out of water’ would be in a sur-
prising state (both emotionally and mathematically).
A fish that frequently forsook water would have high
entropy. Note that both surprise and entropy depend
on the agent: what is surprising for one agent (for
example, being out of water) may not be surprising
for another. Biological agents must therefore mini-
mize the long-term average of surprise to ensure that
their sensory entropy remains low. In other words,
biological systems somehow manage to violate the
fluctuation theorem, which generalizes the second law
of thermodynamics9.
The Wellcome Trust Centre
for Neuroimaging,
University College London,
Queen Square, London,
WC1N 3BG, UK.
e‑mail:
k.friston@fil.ion.ucl.ac.uk
doi:10.1038/nrn2787
Published online
13 January 2010
Free energy
An information theory measure
that bounds or limits (by being
greater than) the surprise on
sampling some data, given a
generative model.
Homeostasis
The process whereby an open
or closed system regulates its
internal environment to
maintain its states within
bounds.
Entropy
The average surprise of
outcomes sampled from a
probability distribution or
density. A density with low
entropy means that, on
average, the outcome is
relatively predictable. Entropy
is therefore a measure of
uncertainty.
The free-energy principle:
a unified brain theory?
Karl Friston
Abstract | A free-energy principle has been proposed recently that accounts for action,
perception and learning. This Review looks at some key brain theories in the biological (for
example, neural Darwinism) and physical (for example, information theory and optimal
control theory) sciences from the free-energy perspective. Crucially, one key theme runs
through each of these theories — optimization. Furthermore, if we look closely at what is
optimized, the same quantity keeps emerging, namely value (expected reward, expected
utility) or its complement, surprise (prediction error, expected cost). This is the quantity that
is optimized under the free-energy principle, which suggests that several global brain
theories might be unified within a free-energy framework.
REVIEWS
NATURE R EVIEWS
|
NEUROSCIENCE VOLUME 11
|
FEBRUARY 2010
|
127
© 20 Macmillan Publishers Limited. All rights reserved10
Nature Reviews | Neuroscience
a
b
Sensations
s
~ = g(x
~, ϑ) + z
~
Action or control signals
a = arg min F(s
~, μ)
Internal states
μ = arg min F(s
~, μ)
External states
˙
x
~ = f(x
~, a, ϑ) + w
~
Environment Agent
Free-energy bound on surprise
F = −<ln p(s
~, ϑ | m)>q + <ln q(ϑ | μ)>q
Action minimizes prediction errors
F = D(q(ϑ | μ) || p(ϑ)) − <ln p(s
~(a) | ϑ, m)>q
a = arg max Accuracy
Perception optimizes predictions
F = D(q(ϑ | μ) || p(ϑ | s
~)) − ln p(s
~ | m)
μ = arg max Divergence
Surprise
(Surprisal or self information.)
The negative log-probability of
an outcome. An improbable
outcome (for example, water
flowing uphill) is therefore
surprising.
Fluctuation theorem
(A term from statistical
mechanics.) Deals with the
probability that the entropy
of a system that is far from the
thermodynamic equilibrium
will increase or decrease over
a given amount of time. It
states that the probability of
the entropy decreasing
becomes exponentially smaller
with time.
Attractor
A set to which a dynamical
system evolves after a long
enough time. Points that
get close to the attractor
remain close, even under
small perturbations.
Kullback-Leibler divergence
(Or information divergence,
information gain or cross
entropy.) A non-commutative
measure of the non-negative
difference between two
probability distributions.
Recognition density
(Or ‘approximating conditional
density’.) An approximate
probability distribution of the
causes of data (for example,
sensory input). It is the product
of inference or inverting a
generative model.
In short, the long-term (distal) imperative — of main-
taining states within physiological bounds — translates
into a short-term (proximal) avoidance of surprise.
Surprise here relates not just to the current state, which
cannot be changed, but also to movement from one state
to another, which can change. This motion can be com-
plicated and itinerant (wandering) provided that it revis-
its a small set of states, called a global random attractor10,
that are compatible with survival (for example, driving a
car within a small margin of error). It is this motion that
the free-energy principle optimizes.
So far, all we have said is that biological agents must
avoid surprises to ensure that their states remain within
physiological bounds (see Supplementary information S1
(box) for a more formal argument). But how do they
do this? A system cannot know whether its sensations
are surprising and could not avoid them even if it did
know. This is where free energy comes in: free energy is
an upper bound on surprise, which means that if agents
minimize free energy, they implicitly minimize surprise.
Crucially, free energy can be evaluated because it is a
function of two things to which the agent has access: its
sensory states and a recognition density that is encoded
by its internal states (for example, neuronal activity
and connection strengths). The recognition density is a
probabilistic representation of what caused a particular
sensation.
This (variational) free-energy construct was
introduced into statistical physics to convert difficult
probability-density integration problems into eas-
ier optimization problems11. It is an information
theoretic quantity (like surprise), as opposed to a
thermo dynamic quantity. Variational free energy has
been exploited in machine learning and statistics to
solve many inference and learning problems12–14. In this
setting, surprise is called the (negative) model evidence.
This means that minimizing surprise is the same as
maximizing the sensory evidence for an agent’s exist-
ence, if we regard the agent as a model of its world. In
the present context, free energy provides the answer to
Box 1 | The free-energy principle
Part a of the figure shows the dependencies among the
quantities that define free energy. These include the
internal states of the brain μ(t) and quantities describing its
exchange with the environment: sensory signals (and their
motion) s˜(t) = [s,s,s]T plus action a(t). The environment
is described by equations of motion, which specify the
trajectory of its hidden states. The causes ϑ ⊃{, θ, γ} of
sensory input comprise hidden states (t), parameters θ
and precisions γ controlling the amplitude of the random
fluctuations (t) and (t). Internal brain states and action
minimize free energy F(,μ), which is a function of sensory
input and a probabilistic representation q(ϑ|μ) of its causes.
This representation is called the recognition density and is
encoded by internal states μ.
The free energy depends on two probability densities:
the recognition density q(ϑ|μ) and one that generates
sensory samples and their causes, p(,ϑ|m). The latter
represents a probabilistic generative model (denoted by
m), the form of which is entailed by the agent or brain.
Part b of the figure provides alternative expressions for the
free energy to show what its minimization entails: action
can reduce free energy only by increasing accuracy (that is,
selectively sampling data that are predicted). Conversely,
optimizing brain states makes the representation an
approximate conditional density on the causes of sensory
input. This enables action to avoid surprising sensory
encounters. A more formal description is provided below.
Optimizing the sufficient statistics (representations)
Optimizing the recognition density makes it a posterior or conditional density on the causes of sensory data: this can be
seen by expressing the free energy as surprise –In p(,| m) plus a Kullback-Leibler divergence between the recognition and
conditional densities (encoded by the ‘internal states’ in the figure). Because this difference is always positive, minimizing
free energy makes the recognition density an approximate posterior probability. This means the agent implicitly infers or
represents the causes of its sensory samples in a Bayes-optimal fashion. At the same time, the free energy becomes a tight
bound on surprise, which is minimized through action.
Optimizing action
Acting on the environment by minimizing free energy enforces a sampling of sensory data that is consistent with the
current representation. This can be seen with a second rearrangement of the free energy as a mixture of accuracy and
complexity. Crucially, action can only affect accuracy (encoded by the ‘external states’ in the figure). This means that
the brain will reconfigure its sensory epithelia to sample inputs that are predicted by the recognition density — in other
words, to minimize prediction error.
REVIEWS
128
|
FEBRUARY 2010
|
VOLUME 11 www.nature.com/reviews/neuro
© 20 Macmillan Publishers Limited. All rights reserved10
Generative model
A probabilistic model (joint
density) of the dependencies
between causes and
consequences (data), from
which samples can be
generated. It is usually
specified in terms of the
likelihood of data, given their
causes (parameters of a model)
and priors on the causes.
Conditional density
(Or posterior density.) The
probability distribution of
causes or model parameters,
given some data; that is, a
probabilistic mapping from
observed data to causes.
Prior
The probability distribution or
density of the causes of data
that encodes beliefs about
those causes before observing
the data.
Bayesian surprise
A measure of salience based
on the Kullback-Leibler
divergence between the
recognition density (which
encodes posterior beliefs) and
the prior density. It
measures the information that
can be recognized in the data.
Bayesian brain hypothesis
The idea that the brain uses
internal probabilistic
(generative) models to update
posterior beliefs, using sensory
information, in an
(approximately) Bayes-optimal
fashion.
Analysis by synthesis
Any strategy (in speech coding)
in which the parameters of a
signal coder are evaluated by
decoding (synthesizing) the
signal and comparing it with
the original input signal.
Epistemological automata
Possibly the first theory for why
top-down influences (mediated
by backward connections in
the brain) might be important
in perception and cognition.
Empirical prior
A prior induced by hierarchical
models; empirical priors
provide constraints on the
recognition density in the usual
way but depend on the data.
a fundamental question: how do self-organizing adap-
tive systems avoid surprising states? They can do this by
minimizing their free energy. So what does this involve?
Implications: action and perception. Agents can
suppress free energy by changing the two things it depends
on: they can change sensory input by acting on the world
or they can change their recognition density by chang-
ing their internal states. This distinction maps nicely
onto action and perception (BOX 1). One can see what this
means in more detail by considering three mathematically
equivalent formulations of free energy (see Supplementary
information S2 (box) for a mathematical treatment).
The first formulation expresses free energy as energy
minus entropy. This formulation is important for three
reasons. First, it connects the concept of free energy as
used in information theory with concepts used in sta-
tistical thermodynamics. Second, it shows that the free
energy can be evaluated by an agent because the energy
is the surprise about the joint occurrence of sensations
and their perceived causes, whereas the entropy is sim-
ply that of the agents own recognition density. Third, it
shows that free energy rests on a generative model of the
world, which is expressed in terms of the probability of a
sensation and its causes occurring together. This means
that an agent must have an implicit generative model of
how causes conspire to produce sensory data. It is this
model that defines both the nature of the agent and the
quality of the free-energy bound on surprise.
The second formulation expresses free energy as
surprise plus a divergence term. The (perceptual) diver-
gence is just the difference between the recognition den-
sity and the conditional density (or posterior density) of the
causes of a sensation, given the sensory signals. This con-
ditional density represents the best possible guess about
the true causes. The difference between the two densities
is always non-negative and free energy is therefore an
upper bound on surprise. Thus, minimizing free energy
by changing the recognition density (without changing
sensory data) reduces the perceptual divergence, so that
the recognition density becomes the conditional density
and the free energy becomes surprise.
The third formulation expresses free energy as com-
plexity minus accuracy, using terms from the model
comparison literature. Complexity is the difference
between the recognition density and the prior density
on causes; it is also known as Bayesian surprise15 and is the
difference between the prior density — which encodes
beliefs about the state of the world before sensory data are
assimilated — and posterior beliefs, which are encoded
by the recognition density. Accuracy is simply the sur-
prise about sensations that are expected under the recog-
nition density. This formulation shows that minimizing
free energy by changing sensory data (without changing
the recognition density) must increase the accuracy of
an agent’s predictions. In short, the agent will selectively
sample the sensory inputs that it expects. This is known
as active inference16. An intuitive example of this process
(when it is raised into consciousness) would be feeling
our way in darkness: we anticipate what we might touch
next and then try to confirm those expectations.
In summary, the free energy rests on a model of how
sensory data are generated and on a recognition density
on the model’s parameters (that is, sensory causes). Free
energy can be reduced only by changing the recognition
density to change conditional expectations about what is
sampled or by changing sensory samples (that is, sensory
input) so that they conform to expectations. In what fol-
lows, I consider these implications in light of some key
theories about the brain.
The Bayesian brain hypothesis
The Bayesian brain hypothesis17 uses Bayesian probability
theory to formulate perception as a constructive process
based on internal or generative models. The underlying
idea is that the brain has a model of the world18–22 that
it tries to optimize using sensory inputs23–28. This idea is
related to analysis by synthesis20 and epistemological autom-
ata19. In this view, the brain is an inference machine that
actively predicts and explains its sensations18,22,25. Central
to this hypothesis is a probabilistic model that can gener-
ate predictions, against which sensory samples are tested
to update beliefs about their causes. This generative
model is decomposed into a likelihood (the probability of
sensory data, given their causes) and a prior (the a priori
probability of those causes). Perception then becomes the
process of inverting the likelihood model (mapping from
causes to sensations) to access the posterior probability of
the causes, given sensory data (mapping from sensations
to causes). This inversion is the same as minimizing the
difference between the recognition and posterior densi-
ties to suppress free energy. Indeed, the free-energy for-
mulation was developed to finesse the difficult problem
of exact inference by converting it into an easier optimi-
zation problem1114. This has furnished some powerful
approximation techniques for model identification and
comparison (for example, variational Bayes or ensemble
learning29). There are many interesting issues that attend
the Bayesian brain hypothesis, which can be illuminated
by the free-energy principle; we will focus on two.
The first is the form of the generative model and
how it manifests in the brain. One criticism of Bayesian
treatments is that they ignore the question of how prior
beliefs, which are necessary for inference, are formed27.
However, this criticism dissolves with hierarchical
generative models, in which the priors themselves are
optimized26,28. In hierarchical models, causes in one
level generate subordinate causes in a lower level; sen-
sory data per se are generated at the lowest level (BOX 2).
Minimizing the free energy effectively optimizes empiri-
cal priors (that is, the probability of causes at one level,
given those in the level above). Crucially, because empir-
ical priors are linked hierarchically, they are informed
by sensory data, enabling the brain to optimize its prior
expectations online. This optimization makes every level
in the hierarchy accountable to the others, furnishing an
internally consistent representation of sensory causes at
multiple levels of description. Not only do hierarchical
models have a key role in statistics (for example, ran-
dom effects and parametric empirical Bayes models30,31),
they may also be used by the brain, given the hierarchical
arrangement of cortical sensory areas32–34.
REVIEWS
NATURE R EVIEWS
|
NEUROSCIENCE VOLUME 11
|
FEBRUARY 2010
|
129
© 20 Macmillan Publishers Limited. All rights reserved10
Nature Reviews | Neuroscience
Sensory
input ξv
(1)
ξx
(1)
˙μv
(i)=Dμv
(i)(i + 1)(i)(i)
− (vε)Tξξ
v
˙μx
(i)=Dμx
(i)(i)(i)
− (xε)Tξ
μθij = −θijεTξ�μγi = ½tr(γi
Π(ξξTΠ(μγ)))
ξv
(2)
ξv
(3)
ξx
(2)
Lower cortical areas Higher cortical areas
Synaptic plasticity Synaptic gain
μx
(1)
μv
(1)
μx
(2)
μv
(2)
s
~(t)
Forward:
prediction
error
Backward:
predictions
ξv
(i)=Πv
(i)=Πv
(i)(i)
εv
(i)(μvg(μ))
(i – 1)
ξx
(i)=Πx
(i)=Πx
(i)(i)
εx
(i)(Dμxf(μ))
(i )
The second issue is the form of the recognition den-
sity that is encoded by physical attributes of the brain,
such as synaptic activity, efficacy and gain. In general,
any density is encoded by its sufficient statistics (for exam-
ple, the mean and variance of a Gaussian form). The way
the brain encodes these statistics places important con-
straints on the sorts of schemes that underlie recognition:
they range from free-form schemes (for example, particle
filtering26 and probabilistic population codes35–38),
which use a vast number of sufficient statistics, to sim-
pler forms, which make stronger assumptions about
the shape of the recognition density, so that it can be
encoded with a small number of sufficient statistics. The
simplest assumed form is Gaussian, which requires only
the conditional mean or expectation — this is known
as the Laplace assumption39, under which the free energy
is just the difference between the model’s predictions
and the sensations or representations that are predicted.
Minimizing free energy then corresponds to explaining
away prediction errors. This is known as predictive coding
and has become a popular framework for understand-
ing neuronal message passing among different levels of
cortical hierarchies40. In this scheme, prediction error
units compare conditional expectations with top-down
predictions to elaborate a prediction error. This predic-
tion error is passed forward to drive the units in the
level above that encode conditional expectations which
optimize top-down predictions to explain away (reduce)
prediction error in the level below. Here, explaining
away just means countering excitatory bottom-up
inputs to a prediction error neuron with inhibitory syn-
aptic inputs that are driven by top-down predictions
(see BOX 2 and REFS 41,42 for detailed discussion). The
reciprocal exchange of bottom-up prediction errors and
top-down predictions proceeds until prediction error
is minimized at all levels and conditional expectations
are optimized. This scheme has been invoked to explain
many features of early visual responses40,43 and provides
a plausible account of repetition suppression and mis-
match responses in electrophysiology44. FIGURE 1 pro-
vides an example of perceptual categorization that uses
this scheme.
Message passing of this sort is consistent with func-
tional asymmetries in real cortical hierarchies45, where
forward connections (which convey prediction errors)
are driving and backwards connections (which model
the nonlinear generation of sensory input) have both
driving and modulatory characteristics46. This asym-
metrical message passing is also a characteristic feature
of adaptive resonance theory47,48, which has formal simi-
larities to predictive coding.
In summary, the theme underlying the Bayesian brain
and predictive coding is that the brain is an inference
engine that is trying to optimize probabilistic representa-
tions of what caused its sensory input. This optimization
can be finessed using a (variational free-energy) bound
on surprise. In short, the free-energy principle entails
the Bayesian brain hypothesis and can be implemented
by the many schemes considered in this field. Almost
invariably, these involve some form of message passing
or belief propagation among brain areas or units. This
Box 2 | Hierarchical message passing in the brain
The figure details a neuronal architecture that optimizes the conditional expectations of
causes in hierarchical models of sensory input. It shows the putative cells of origin of forward
driving connections that convey prediction error (grey arrows) from a lower area (for
example, the lateral geniculate nucleus) to a higher area (for example, V1), and nonlinear
backward connections (black arrows) that construct predictions41. These predictions try to
explain away prediction error in lower levels. In this scheme, the sources of forward and
backward connections are superficial and deep pyramidal cells (upper and lower triangles),
respectively, where state units are black and error units are grey. The equations represent a
gradient descent on free energy using the generative model below. The two upper equations
describe the formation of prediction error encoded by error units, and the two lower
equations represent recognition dynamics, using a gradient descent on free energy.
Generative models in the brain
To evaluate free energy one needs a generative model of how the sensorium is caused.
Such models p(,ϑ) = p( | ϑ) p(ϑ) combine the likelihood p( | ϑ) of getting some data given
their causes and the prior beliefs about these causes, p(ϑ). The brain has to explain
complicated dynamics on continuous states with hierarchical or deep causal structure
and may use models with the following form
Nature Reviews | Neuroscience
Z
·  HZX θY
U IZX θ\
Z
· K HZKXK θKYK
XKs IZKXK θK\K
……
Here, g(i) and f(i) are continuous nonlinear functions of (hidden and causal) states, with
parameters θ(i). The random fluctuations z(t)(i) and w(t)(i) play the part of observation
noise at the sensory level and state noise at higher levels. Causal states v(t)(i) link
hierarchical levels, where the output of one level provides input to the next. Hidden
states x(t)(i) link dynamics over time and endow the model with memory.
Gaussian assumptions about the random fluctuations specify the likelihood
and Gaussian assumptions about state noise furnish empirical priors in terms of
predicted motion. These assumptions are encoded by their precision (or inverse
variance), П(i)(γ), which are functions of precision parameters γ.
Recognition dynamics and prediction error
If we assume that neuronal activity encodes the conditional expectation of states, then
recognition can be formulated as a gradient descent on free energy. Under Gaussian
assumptions, these recognition dynamics can be expressed compactly in terms
of precision-weighted prediction errors ξ(i) = П(i)(ε)(i) on the causal states and motion of
hidden states. The ensuing equations (see the figure) suggest two neuronal populations
that exchange messages: causal or hidden-state units encoding expected states and
error units encoding prediction error. Under hierarchical models, error units receive
messages from the state units in the same level and the level above, whereas state units
are driven by error units in the same level and the level below. These provide bottom-up
messages that drive conditional expectations μ(i) towards better predictions, which
explain away prediction error. These top-down predictions correspond to g(μ(i)) and f(μ(i)).
This scheme suggests that the only connections that link levels are forward connections
conveying prediction error to state units and reciprocal backward connections that
mediate predictions. See REFS 42,130 for details. Figure is modified from REF. 42.
REVIEWS
130
|
FEBRUARY 2010
|
VOLUME 11 www.nature.com/reviews/neuro
© 20 Macmillan Publishers Limited. All rights reserved10
Nature Reviews | Neuroscience
a Perceptual inference
b Perceptual categorization
c
0 0.2 0.4 0.6 0.8 1
–20
–10
0
10
20
30
40
50
5,000
4,000
3,000
2,000
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
a
b
c
Time (s)
Time (s)
Estimated causes
Frequency (Hz)
µv1
µv1
10 15 20 25 30 35
1
1.5
2
2.5
3
3.5
a
b
c
v2
v1
Song a Song b Song c
v = v2
v1
Vocal centre Syrinx Sonogram
˙x = f(x, v) = v1x1 − 2x3x1x2
18x2 − 18x1
2x1x2v2x3
allows us to connect the free-energy principle to another
principled approach to sensory processing, namely
information theory.
The principle of efficient coding
The principle of efficient coding suggests that the brain
optimizes the mutual information (that is, the mutual
predictability) between the sensorium and its internal
representation, under constraints on the efficiency of
those representations. This line of thinking was articu-
lated by Barlow49 in terms of a redundancy reduction
principle (or principle of efficient coding) and formal-
ized later in terms of the infomax principle50. It has been
applied in machine learning51, leading to methods
like independent component analysis52, and in neuro-
biology, contributing to an understanding of the nature
of neuronal responses53–56. This principle is extremely
effective in predicting the empirical characteristics of
classical receptive fields53 and provides a principled
explanation for sparse coding55 and the segregation of
processing streams in visual hierarchies57. It has been
extended to cover dynamics and motion trajectories58,59
and even used to infer the metabolic constraints on neu-
ronal processing60.
At its simplest, the infomax principle says that
neuronal activity should encode sensory information in
an efficient and parsimonious fashion. It considers the
mapping between one set of variables (sensory states)
and another (variables representing those states). At
first glance, this seems to preclude a probabilistic repre-
sentation, because this would involve mapping between
sensory states and a probability density. However, the
infomax principle can be applied to the sufficient sta-
tistics of a recognition density. In this context, the info-
max principle becomes a special case of the free-energy
principle, which arises when we ignore uncertainty
in probabilistic representations (and when there is no
action); see Supplementary information S3 (box) for
mathematical details). This is easy to see by noting that
sensory signals are generated by causes. This means that it
is sufficient to represent the causes to predict these
signals. More formally, the infomax principle can be
understood in terms of the decomposition of free energy
into complexity and accuracy: mutual information is
optimized when conditional expectations maximize
accuracy (or minimize prediction error), and efficiency
is assured by minimizing complexity. This ensures that
no excessive parameters are applied in the generative
model and leads to a parsimonious representation of
sensory data that conforms to prior constraints on their
causes. Interestingly, advanced model-optimization
techniques use free-energy optimization to eliminate
redundant model parameters61, suggesting that free-
energy optimization might provide a nice explanation
for the synaptic pruning and homeostasis that take place
in the brain during neurodevelopment62 and sleep63.
The infomax principle pertains to a forward mapping
from sensory input to representations. How does this
square with optimizing generative models, which map
from causes to sensory inputs? These perspectives can be
reconciled by noting that all recognition schemes based
Figure 1 | Birdsongs and perceptual categorization. a | The generative model of
birdsong used in this simulation comprises a Lorenz attractor with two control parameters
(or causal states) (v1,v2), which, in turn, delivers two control parameters (not shown) to a
synthetic syrinx to produce ‘chirps’ that were modulated in amplitude and frequency (an
example is shown as a sonogram). The chirps were then presented as a stimulus to a
synthetic bird to see whether it could infer the underlying causal states and thereby
categorize the song. This entails minimizing free energy by changing the internal
representation (v1,v2) of the control parameters. Examples of this perceptual inference or
categorization are shown below. b | Three simulated songs are shown in sonogram format.
Each comprises a series of chirps, the frequency and number of which fall progressively
from song a to song c, as a causal state (known as the Raleigh number; v1 in part a) is
decreased. c | The graph on the left depicts the conditional expectations (v1,v2) of the
causal states, shown as a function of peristimulus time for the three songs. It shows that
the causes are identified after around 600 ms with high conditional precision (90%
confidence intervals are shown in grey). The graph on the right shows the conditional
density on the causes shortly before the end of the peristimulus time (that is, the dotted
line in the left panel). The blue dots correspond to conditional expectations and the grey
areas correspond to the 90% conditional confidence regions. Note that these encompass
the true values (red dots) of (v1,v2) that were used to generate the songs. These results
illustrate the nature of perceptual categorization under the inference scheme in BOX 2:
here, recognition corresponds to mapping from a continuously changing and chaotic
sensory input to a fixed point in perceptual space. Figure is reproduced, with permission,
from REF. 130 © (2009) Elsevier.
REVIEWS
NATURE R EVIEWS
|
NEUROSCIENCE VOLUME 11
|
FEBRUARY 2010
|
131
© 20 Macmillan Publishers Limited. All rights reserved10
Sufficient statistics
Quantities that are sufficient to
parameterize a probability
density (for example, mean and
covariance of a Gaussian
density).
Laplace assumption
(Or Laplace approximation or
method.) A saddle-point
approximation of the integral
of an exponential function, that
uses a second-order Taylor
expansion. When the function
is a probability density, the
implicit assumption is that
the density is approximately
Gaussian.
Predictive coding
A tool used in signal processing
for representing a signal using
a linear predictive (generative)
model. It is a powerful speech
analysis technique and was
first considered in vision to
explain lateral interactions in
the retina.
Infomax
An optimization principle for
neural networks (or functions)
that map inputs to outputs. It
says that the mapping should
maximize the Shannon mutual
information between the inputs
and outputs, subject to
constraints and/or noise
processes.
Stochastic
Governed by random effects.
Biased competition
An attentional effect mediated
by competitive interactions
among neurons representing
visual stimuli; these
interactions can be biased in
favour of behaviourally relevant
stimuli by both spatial and
non-spatial and both
bottom-up and top-down
processes.
on infomax can be cast as optimizing the parameters of a
generative model64. For example, in sparse coding mod-
els55, the implicit priors posit independent causes that
are sampled from a heavy-tailed or sparse distribution42.
The fact that these models predict empirically observed
receptive fields so well suggests that we are endowed
with (or acquire) prior expectations that the causes of
our sensations are largely independent and sparse.
In summary, the principle of efficient coding says
that the brain should optimize the mutual information
between its sensory signals and some parsimonious
neuronal representations. This is the same as optimizing
the parameters of a generative model to maximize the
accuracy of predictions, under complexity constraints.
Both are mandated by the free-energy principle, which
can be regarded as a probabilistic generalization of the
infomax principle. We now turn to more biologically
inspired ideas about brain function that focus on neu-
ronal dynamics and plasticity. This takes us deeper into
neurobiological mechanisms and the implementation of
the theoretical principles outlined above.
The cell assembly and correlation theory
The cell assembly theory was proposed by Hebb65 and
entails Hebbian — or associative — plasticity, which is a
cornerstone of use-dependent or experience-dependent
plasticity66, the correlation theory of von de Malsburg67,68
and other formal refinements to Hebbian plasticity
per se69. The cell assembly theory posits that groups of
interconnected neurons are formed through a strength-
ening of synaptic connections that depends on corre-
lated pre- and postsynaptic activity; that is, ‘cells that fire
together wire together’. This enables the brain to distil
statistical regularities from the sensorium. The correla-
tion theory considers the selective enabling of synaptic
efficacy and its plasticity (also known as metaplastic-
ity70) by fast synchronous activity induced by different
perceptual attributes of the same object (for example, a
red bus in motion). This resolves a putative deficiency
of classical plasticity, which cannot ascribe a presynaptic
input to a particular cause (for example, redness) in the
world67. The correlation theory underpins theoretical
treatments of synchronized brain activity and its role in
associating or binding attributes to specific objects or
causes68,71. Another important field that rests on associa-
tive plasticity is the use of attractor networks as models
of memory formation and retrieval72–74. So how do corre-
lations and associative plasticity figure in the free-energy
formulation?
Hitherto, we have considered only inference on states
of the world that cause sensory signals, whereby condi-
tional expectations about states are encoded by synaptic
activity. However, the causes covered by the recognition
density are not restricted to time-varying states (for
example, the motion of an object in the visual field):
they also include time-invariant regularities that endow
the world with causal structure (for example, objects
fall with constant acceleration). These regularities are
parameters of the generative model and have to be
inferred by the brain — in other words, the conditional
expectations of these parameters that may be encoded
by synaptic efficacy (these are θ in BOX 2) have to be
optimized. This corresponds to optimizing connection
strengths in the brain — that is, plasticity that under-
lines learning. So what form would this learning take? It
transpires that a gradient descent on free energy (that is,
changing connections to reduce free energy) is formally
identical to Hebbian plasticity28,42 (BOX 2). This is because
the parameters of the generative model determine how
expected states (synaptic activity) are mixed to form pre-
dictions. Put simply, when the presynaptic predictions
and postsynaptic prediction errors are highly correlated,
the connection strength increases, so that predictions
can suppress prediction errors more efficiently.
In short, the formation of cell assemblies reflects the
encoding of causal regularities. This is just a restate-
ment of cell assembly theory in the context of a specific
implementation (predictive coding) of the free-energy
principle. It should be acknowledged that the learning
rule in predictive coding is really a delta rule, which
rests on Hebbian mechanisms; however, Hebb’s wider
notions of cell assemblies were formulated from a non-
statistical perspective. Modern reformulations suggest
that both inference on states (that is, perception) and
inference on parameters (that is, learning) minimize
free energy (that is, minimize prediction error) and
serve to bound surprising exchanges with the world. So
what about synchronization and the selective enabling
of synapses?
Biased competition and attention
Causal regularities encoded by synaptic efficacy
control the deterministic evolution of states in the world.
However, stochastic (that is, random) fluctuations in
these states play an important part in generating sen-
sory data. Their amplitude is usually represented as pre-
cision (or inverse variance), which encodes the reliability
of prediction errors. Precision is important, especially
in hierarchical schemes, because it controls the relative
influence of bottom-up prediction errors and top-down
predictions. So how is precision encoded in the brain?
In predictive coding, precision modulates the amplitude
of prediction errors (these are γ in BOX 2), so that pre-
diction errors with high precision have a greater impact
on units that encode conditional expectations. This
means that precision corresponds to the synaptic gain of
prediction error units. The most obvious candidates for
controlling gain (and implicitly encoding precision) are
classical neuromodulators like dopamine and acetylcho-
line, which provides a nice link to theories of attention
and uncertainty75–77. Another candidate is fast synchro-
nized presynaptic input that lowers effective postsynaptic
membrane time constants and increases synchronous
gain78. This fits comfortably with the correlation theory
and speaks to recent ideas about the role of synchronous
activity in mediating attentional gain79,80.
In summary, the optimization of expected precision
in terms of synaptic gain links attention to synaptic gain
and synchronization. This link is central to theories of
attentional gain and biased competition80–85, particularly
in the context of neuromodulation86,87. The theories
considered so far have dealt only with perception.
REVIEWS
132
|
FEBRUARY 2010
|
VOLUME 11 www.nature.com/reviews/neuro
© 20 Macmillan Publishers Limited. All rights reserved10
Reentrant signalling
Reciprocal message passing
among neuronal groups.
Reinforcement learning
An area of machine learning
concerned with how an agent
maximizes long-term reward.
Reinforcement learning
algorithms attempt to find a
policy that maps states of the
world to actions performed by
the agent.
Optimal control theory
An optimization method
(based on the calculus of
variations) for deriving an
optimal control law in a
dynamical system. A control
problem includes a cost
function that is a function of
state and control variables.
Bellman equation
(Or dynamic programming
equation.) Named after
Richard Bellman, it is a
necessary condition for
optimality associated with
dynamic programming in
optimal control theory.
However, from the point of view of the free-energy
principle, perception just makes free energy a good
proxy for surprise. To actually reduce surprise we need
to act. In the next section, we retain a focus on cell
assemblies but move to the selection and reinforcement
of stimulus–response links.
Neural Darwinism and value learning
In the theory of neuronal group selection88, the emergence
of neuronal assemblies is considered in the light of selec-
tive pressure. The theory has four elements: epigenetic
mechanisms create a primary repertoire of neuronal
connections, which are refined by experience-dependent
plasticity to produce a secondary repertoire of neuro-
nal groups. These are selected and maintained through
reentrant signalling among neuronal groups. As in cell
assembly theory, plasticity rests on correlated pre- and
postsynaptic activity, but here it is modulated by value.
Value is signalled by ascending neuromodulatory trans-
mitter systems and controls which neuronal groups
are selected and which are not. The beauty of neural
Darwinism is that it nests distinct selective processes
within each other. In other words, it eschews a single unit
of selection and exploits the notion of meta-selection
(the selection of selective mechanisms; for example, see
REF. 89). In this context, (neuronal) value confers evolu-
tionary value (that is, adaptive fitness) by selecting neu-
ronal groups that meditate adaptive stimulus–stimulus
associations and stimulus–response links. The capacity
of value to do this is assured by natural selection, in the
sense that neuronal value systems are themselves subject
to selective pressure.
This theory, particularly value-dependent learning90,
has deep connections with reinforcement learning and
related approaches in engineering (see below), such as
dynamic programming and temporal difference mod-
els91,92. This is because neuronal value systems reinforce
connections to themselves, thereby enabling the brain
to label a sensory state as valuable if, and only if, it leads to
another valuable state. This ensures that agents move
through a succession of states that have acquired value to
access states (rewards) with genetically specified innate
value. In short, the brain maximizes value, which may be
reflected in the discharge of value systems (for example,
dopaminergic systems92–96). So how does this relate to
the optimization of free energy?
The answer is simple: value is inversely proportional
to surprise, in the sense that the probability of a pheno-
type being in a particular state increases with the value
of that state. Furthermore, the evolutionary value of
a phenotype is the negative surprise averaged over all
the states it experiences, which is simply its negative
entropy. Indeed, the whole point of minimizing free
energy (and implicitly entropy) is to ensure that agents
spend most of their time in a small number of valuable
states. This means that free energy is the complement of
value, and its long-term average is the complement of
adaptive fitness (also known as free fitness in evolution-
ary biology97). But how do agents know what is valu-
able? In other words, how does one generation tell the
next which states have value (that is, are unsurprising)?
Value or surprise is determined by the form of an agent’s
generative model and its implicit priors — these specify
the value of sensory states and, crucially, are heritable
through genetic and epigenetic mechanisms. This means
that prior expectations (that is, the primary repertoire)
can prescribe a small number of attractive states with
innate value. In turn, this enables natural selection to
optimize prior expectations and ensure they are con-
sistent with the agent’s phenotype. Put simply, valuable
states are just the states that the agent expects to fre-
quent. These expectations are constrained by the form of
its generative model, which is specified genetically and
fulfilled behaviourally, under active inference.
It is important to appreciate that prior expectations
include not just what will be sampled from the world but
also how the world is sampled. This means that natural
selection may equip agents with the prior expectation
that they will explore their environment until states
with innate value are encountered. We will look at this
more closely in the next section, where priors on motion
through state space are cast in terms of policies in
reinforcement learning.
Both neural Darwinism and the free-energy principle
try to understand somatic changes in an individual in
the context of evolution: neural Darwinism appeals to
selective processes, whereas the free energy formulation
considers the optimization of ensemble or population
dynamics in terms of entropy and surprise. The key
theme that emerges here is that (heritable) prior expecta-
tions can label things as innately valuable (unsurprising);
but how can simply labelling states engender adaptive
behaviour? In the next section, we return to reinforce-
ment learning and related formulations of action that try
to explain adaptive behaviour purely in terms of labels
or cost functions.
Optimal control theory and game theory
Value is central to theories of brain function that are
based on reinforcement learning and optimum con-
trol. The basic notion that underpins these treatments
is that the brain optimizes value, which is expected
reward or utility (or its complement — expected loss
or cost). This is seen in behavioural psychology as rein-
forcement learning98, in computational neuroscience
and machine learning as variants of dynamic program-
ming such as temporal difference learning99101, and in
economics as expected utility theory102. The notion of
an expected reward or cost is crucial here; this is the
cost expected over future states, given a particular policy
that prescribes action or choices. A policy specifies the
states to which an agent will move from any given state
(‘motion through state space in continuous time’). This
policy has to access sparse rewarding states using a cost
function, which only labels states as costly or not. The
problem of how the policy is optimized is formalized
in optimal control theory as the Bellman equation and its
variants99 (see Supplementary information S4 (box)),
which express value as a function of the optimal policy
and a cost function. If one can solve the Bellman equa-
tion, one can associate each sensory state with a value
and optimize the policy by ensuring that the next state
REVIEWS
NATURE R EVIEWS
|
NEUROSCIENCE VOLUME 11
|
FEBRUARY 2010
|
133
© 20 Macmillan Publishers Limited. All rights reserved10
Nature Reviews | Neuroscience
svisual =+ wvisual
J
V
sprop =+ wprop
x2
x1
V = (v1, v2, v3)
J = J1 + J2 = ( j1, j2)
Movement
trajectory
(0, 0)
x1
J1
J2
x2
Jointed arm
a
ξv
(1)
ξv
(1)
ξx
(1)
ξv
(2)
μx
(1)
μv
(1)
Action
Motor
signals
Predictions
Prediction errors
˙a = −aεTξ
Optimal decision theory
(Or game theory.) An area of
applied mathematics
concerned with identifying the
values, uncertainties and other
constraints that determine an
optimal decision.
Gradient ascent
(Or method of steepest
ascent.) A first-order
optimization scheme that finds
a maximum of a function by
changing its arguments in
proportion to the gradient of
the function at the current
value. In short, a hill-climbing
scheme. The opposite scheme
is a gradient descent.
is the most valuable of the available states. In general,
it is impossible to solve the Bellman equation exactly,
but several approximations exist, ranging from simple
Rescorla–Wagner models98 to more comprehensive for-
mulations like Q-learning100. Cost also has a key role in
Bayesian decision theory, in which optimal decisions
minimize expected cost in the context of uncertainty
about outcomes; this is central to optimal decision theory
(game theory) and behavioural economics102–104.
So what does free energy bring to the table? If one
assumes that the optimal policy performs a gradient
ascent on value, then it is easy to show that value is
inversely proportional to surprise (see Supplementary
information S4 (box)). This means that free energy is
(an upper bound on) expected cost, which makes sense
as optimal control theory assumes that action mini-
mizes expected cost, whereas the free-energy principle
states that it minimizes free energy. This is important
because it explains why agents must minimize expected
cost. Furthermore, free energy provides a quantitative
and seamless connection between the cost functions
of reinforcement learning and value in evolutionary
biology. Finally, the dynamical perspective provides a
mechanistic insight into how policies are specified in the
brain: according to the principle of optimality99 cost is the
rate of change of value (see Supplementary information
S4 (box)), which depends on changes in sensory states.
This suggests that optimal policies can be prescribed by
prior expectations about the motion of sensory states.
Put simply, priors induce a fixed-point attractor, and
when the states arrive at the fixed point, value will stop
changing and cost will be minimized. A simple exam-
ple is shown in FIG. 2, in which a cued arm movement
is simulated using only prior expectations that the arm
will be drawn to a fixed point (the target). This figure
illustrates how computational motor control105109 can
be formulated in terms of priors and the suppression of
sensory prediction errors (K.J.F., J. Daunizeau, J. Kilner
and S.J. Kiebel, unpublished observations). More gener-
ally, it shows how rewards and goals can be considered
as prior expectations that an action is obliged to fulfil16
(see also REF. 110). It also suggests how natural selection
could optimize behaviour through the genetic specifi-
cation of inheritable or innate priors that constrain the
learning of empirical priors (BOX 2) and subsequent goal-
directed action.
It should be noted that just expecting to be attracted
to some states may not be sufficient to attain those states.
This is because one may have to approach attractors vicar-
iously through other states (for example, to avoid obsta-
cles) or conform to physical constraints on action. These
are some of the more difficult problems of accessing
distal rewards that reinforcement learning and opti-
mum control contend with. In these circumstances,
an examination of the density dynamics, on which the
free-energy principle is based, suggests that it is sufficient
to keep moving until an a priori attractor is encountered
(see Supplementary information S5 (box)). This entails
destroying unexpected (costly) fixed points in the envi-
ronment by making them unstable (like shifting to a new
position when sitting uncomfortably). Mathematically,
this means adopting a policy that ensures a positive
divergence in costly states (intuitively, this is like being
pushed through a liquid with negative viscosity or
friction). See FIG. 3 for a solution to the classical
mountain car problem using a simple prior that induces
this sort of policy. This prior is on motion through state
space (that is, changes in states) and enforces exploration
until an attractive state is found. Priors of this sort may
provide a principled way to understand the exploration–
exploitation trade-off111113 and related issues in evolu-
tionary biology114. The implicit use of priors to induce
dynamical instability also provides a key connection
to dynamical systems theory approaches to the brain
that emphasize the importance of itinerant dynamics,
metastability, self-organized criticality and winner-
less competition115–123. These dynamical phenomena
have a key role in synergetic and autopoietic accounts of
adaptive behaviour5,124,125.
Figure 2 | A demonstration of cued reaching movements. The lower right part of the
figure shows a motor plant, comprising a two-jointed arm with two hidden states, each of
which corresponds to a particular angular position of the two joints; the current position
of the finger (red circle) is the sum of the vectors describing the location of each joint.
Here, causal states in the world are the position and brightness of the target (green
circle). The arm obeys Newtonian mechanics, specified in terms of angular inertia and
friction. The left part of the figure illustrates that the brain senses hidden states directly
in terms of proprioceptive input (Sprop) that signals the angular positions (x1,x2) of the
joints and indirectly through seeing the location of the finger in space (J1,J2). In addition,
through visual input (Svisual) the agent senses the target location (v1,v2) and brightness (v3).
Sensory prediction errors are passed to higher brain levels to optimize the conditional
expectations of hidden states (that is, the angular position of the joints) and causal (that
is, target) states. The ensuing predictions are sent back to suppress sensory prediction
errors. At the same time, sensory prediction errors are also trying to suppress themselves
by changing sensory input through action. The grey and black lines denote reciprocal
message passing among neuronal populations that encode prediction error and
conditional expectations; this architecture is the same as that depicted in BOX 2. The
blue lines represent descending motor control signals from sensory prediction-error
units. The agent’s generative model included priors on the motion of hidden states that
effectively engage an invisible elastic band between the finger and target (when the
target is illuminated). This induces a prior expectation that the finger will be drawn to
the target, when cued appropriately. The insert shows the ensuing movement trajectory
caused by action. The red circles indicate the initial and final positions of the finger,
which reaches the target (green circle) quickly and smoothly; the blue line is the
simulated trajectory.
REVIEWS
134
|
FEBRUARY 2010
|
VOLUME 11 www.nature.com/reviews/neuro
© 20 Macmillan Publishers Limited. All rights reserved10
Nature Reviews | Neuroscience
f ==
˙x
˙x
xϕ 18 x + σ(a)
x
Equations of motion
-2 -1 0 12
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Position (x)
The mountain car problem
20 40 60 80 100120
–3
–2
–1
0
1
2
3
Time (seconds)
Time (seconds)
20 40 60 80 100120
–5
0
0
0
5
10
15
20
25
30
Conditional expectations
Height
ϕ(x)
μ(t)x
c(t)
a(t)
Estimated states
Control signal
Position (x)
–2 –1 012
–30
–25
–20
–15
–10
–5
0
5
Loss functions (priors)
Force
c(x)
Position (x)
–2 –1 012
–2
–1
0
1
2
x
Velocity
Trajectories Action
ab
Principle of optimality
An optimal policy has
the property that whatever the
initial state and initial decision,
the remaining decisions must
constitute an optimal policy
with regard to the state
resulting from the first decision.
Exploration–exploitation
trade-off
Involves a balance between
exploration (of uncharted
territory) and exploitation (of
current knowledge). In
reinforcement learning, it has
been studied mainly through
the multi-armed bandit
problem.
Dynamical systems theory
An area of applied
mathematics that describes
the behaviour of complex
(possibly chaotic) dynamical
systems as described by
differential or difference
equations.
Synergetics
Concerns the self-organization
of patterns and structures in
open systems far from
thermodynamic equilibrium. It
rests on the order parameter
concept, which was generalized
by Haken to the enslaving
principle: that is, the dynamics
of fast-relaxing (stable) modes
are completely determined by
the ‘slow’ dynamics of order
parameters (the amplitudes of
unstable modes).
Autopoietic
Referring to the fundamental
dialectic between structure
and function.
Helmholtzian
Refers to a device or scheme
that uses a generative model to
furnish a recognition density
and learns hidden structures in
data by optimizing the
parameters of generative
models.
In summary, optimal control and decision (game)
theory start with the notion of cost or utility and try to
construct value functions of states, which subsequently
guide action. The free-energy formulation starts with
a free-energy bound on the value of states, which is
specified by priors on the motion of hidden environ-
mental states. These priors can incorporate any cost
function to ensure that costly states are avoided. States
with minimum cost can be set (by learning or evolu-
tion) in terms of prior expectations about motion and
the attractors that ensue. In this view, the problem of
finding sparse rewards in the environment is nature’s
solution to the problem of how to minimize the entropy
(average surprise or free energy) of an agent’s states: by
ensuring they occupy a small set of attracting (that is,
rewarding) states.
Conclusions and future directions
Although contrived to highlight commonalities, this
Review suggests that many global theories of brain
function can be united under a Helmholtzian percep-
tive of the brain as a generative model of the world it
inhabits18,20,21,25 (FIG. 4); notable examples include the
integration of the Bayesian brain and computational
motor control theory, the objective functions shared
by predictive coding and the infomax principle,
hierarchical inference and theories of attention, the
embedding of perception in natural selection and
the link between optimum control and more exotic
phenomena in dynamical systems theory. The constant
theme in all these theories is that the brain optimizes
a (free-energy) bound on surprise or its complement,
value. This manifests as perception (so as to change
Figure 3 | Solving the mountain car problem with prior expectations. a | How paradoxical but adaptive behaviour (for
example, moving away from a target to ensure that it is secured later) emerges from simple priors on the motion of hidden
states in the world. Shown is the landscape or potential energy function (with a minimum at position x = –0.5) that exerts
forces on a mountain car. The car is shown at the target position on the hill at x =1, indicated by the red circle. The equations
of motion of the car are shown below the plot. Crucially, at x = 0 the force on the car cannot be overcome by the agent,
because a squashing function –1≤σ≤1 is applied to action to prevent it being greater than 1. This means that the agent can
access the target only by starting halfway up the left hill to gain enough momentum to carry it up the other side. b | The
results of active inference under priors that destabilize fixed points outside the target domain. The priors are encoded in a
cost function c(x) (top left), which acts like negative friction. When ‘friction’ is negative the car expects to go faster (see
Supplementary information S5 (box) for details). The inferred hidden states (upper right: position in blue, velocity in green
and negative dissipation in red) show that the car explores its landscape until it encounters the target, and that friction then
increases (that is, cost decreases) dramatically to prevent the car from escaping the target (by falling down the hill). The
ensuing trajectory is shown in blue (bottom left). The paler lines provide exemplar trajectories from other trials, with
different starting positions. In the real world, friction is constant. However, the car ‘expects’ friction to change as it changes
position, thus enforcing exploration or exploitation. These expectations are fulfilled by action (lower right).
REVIEWS
NATURE R EVIEWS
|
NEUROSCIENCE VOLUME 11
|
FEBRUARY 2010
|
135
© 20 Macmillan Publishers Limited. All rights reserved10
Nature Reviews | Neuroscience
˙μv
(i)=Dμv
(i)(i)(i)
vεTξ(i + 1)
ξv
μθij = −θijεTξ
˙a = −aεTξ
Infomax and the redundancy
minimization principle
Maximization of the mutual
information between sensations
and representations
Probabilistic neuronal coding
Encoding a recognition density
in terms of conditional
expectations and uncertainty
The Bayesian brain hypothesis
Minimizing the difference between a
recognition density and the conditional
density on sensory causes
Computational motor control
Minimization of sensory
prediction errors
Predictive coding and hierarchical inference
Minimization of prediction error
with recurrent message passing
Perceptual learning and memory
Optimization of synaptic efficacy
to represent causal structure
in the sensorium
Associative plasticity
Optimization of synaptic efficacy
Optimal control and value learning
Optimization of a free-energy
bound on surprise or value
Model selection and evolution
The free-energy principle
Optimizing the agent’s model and
priors through neurodevelopment
and natural selection
Minimization of the free energy of
sensations and the representation
of their causes
Attention and biased competition
Optimization of synaptic gain
representing the precision
(salience) of predictions
m = arg min dtF
a, μ, m = arg min F (s
~, μ | m)
a, μ = arg max V (s
~ | m)
μ = arg max {I (s
~, μ ) H(μ)}
μθ = arg min dtF
μγ = arg min dtF
μ = arg min DKL(q(ϑ) || (p(ϑ | s
~))
q(ϑ ) = N ( μ, Σ)
predictions) or action (so as to change the sensations
that are predicted). Crucially, these predictions depend
on prior expectations (that furnish policies), which
are optimized at different (somatic and evolutionary)
timescales and define what is valuable.
What does the free-energy principle portend for the
future? If its main contribution is to integrate estab-
lished theories, then the answer is probably ‘not a lot’.
Conversely, it may provide a framework in which cur-
rent debates could be resolved, for example whether
dopamine encodes reward prediction error or sur-
prise126,127 — this is particularly important for under-
standing conditions like addiction, Parkinson’s disease
and schizophrenia. Indeed, the free-energy formulation
has already been used to explain the positive symptoms
of schizophrenia in terms of false inference128. The free-
energy formulation could also provide new approaches
Figure 4 | The free-energy principle and other theories. Some of the theoretical constructs considered in this Review
and how they relate to the free-energy principle (centre). The variables are described in BOXES 1,2 and a full explanation
of the equations can be found in the Supplementary information S1–S4 (boxes).
to old problems that might call for a reappraisal of
conventional notions, particularly in reinforcement
learning and motor control.
If the arguments underlying the free-energy principle
hold, then the real challenge is to understand how it
manifests in the brain. This speaks to a greater appre-
ciation of hierarchical message passing41, the func-
tional role of specific neurons and microcircuits and
the dynamics they support (for example, what is the
relationship between predictive coding, attention
and dynamic co ordination in the brain?129). Beyond
neuroscience, many exciting applications in engineering,
robotics, embodied cognition and evolutionary biology
suggest themselves; although fanciful, it is not difficult to
imagine building little free-energy machines that garner
and model sensory information (like our children) to
maximize the evidence for their own existence.
1. Huang, G. Is this a unified theory of the brain?
New Scientist 2658, 30–33 (2008).
2. Friston K., Kilner, J. & Harrison, L. A free energy
principle for the brain. J. Physiol. Paris 100, 70–87
(2006).
An overview of the free-energy principle that
describes its motivation and relationship to
generative models and predictive coding. This
paper focuses on perception and the
neurobiological infrastructures involved.
3. Ashby, W. R. Principles of the self-organising dynamic
system. J. Gen. Psychol. 37, 125–128 (1947).
4. Nicolis, G. & Prigogine, I. Self‑Organisation in Non‑
Equilibrium Systems (Wiley, New York, 1977).
5. Haken, H. Synergistics: an Introduction. Non‑
Equilibrium Phase Transition and Self‑Organisation in
Physics, Chemistry and Biology 3rd edn (Springer,
New York, 1983).
6. Kauffman, S. The Origins of Order: Self‑Organization
and Selection in Evolution (Oxford Univ. Press, Oxford,
1993).
7. Bernard, C. Lectures on the Phenomena Common
to Animals and Plants (Thomas, Springfield,
1974).
REVIEWS
136
|
FEBRUARY 2010
|
VOLUME 11 www.nature.com/reviews/neuro
© 20 Macmillan Publishers Limited. All rights reserved10
8. Applebaum, D. Probability and Information: an
Integrated Approach (Cambridge Univ. Press,
Cambridge, UK, 2008).
9. Evans, D. J. A non-equilibrium free energy theorem
for deterministic systems. Mol. Physics 101,
15551–11554 (2003).
10. Crauel, H. & Flandoli, F. Attractors for random
dynamical systems. Probab. Theory Relat. Fields 100,
365–393 (1994).
11. Feynman, R. P. Statistical Mechanics: a Set of Lectures
(Benjamin, Reading, Massachusetts, 1972).
12. Hinton, G. E. & von Cramp, D. Keeping neural
networks simple by minimising the description length
of weights. Proc. 6th Annu. ACM Conf. Computational
Learning Theory 5–13 (1993).
13. MacKay. D. J. C. Free-energy minimisation algorithm
for decoding and cryptoanalysis. Electron. Lett. 31,
445–447 (1995).
14. Neal, R. M. & Hinton, G. E. in Learning in Graphical
Models (ed. Jordan, M. I.) 355–368 (Kluwer
Academic, Dordrecht, 1998).
15. Itti, L. & Baldi, P. Bayesian surprise attracts human
attention. Vision Res. 49, 1295–1306 (2009).
16. Friston, K., Daunizeau, J. & Kiebel, S. Active inference
or reinforcement learning? PLoS ONE 4, e6421
(2009).
17. Knill, D. C. & Pouget, A. The Bayesian brain: the role
of uncertainty in neural coding and computation.
Trends Neurosci. 27, 712–719 (2004).
A nice review of Bayesian theories of perception
and sensorimotor control. Its focus is on Bayes
optimality in the brain and the implicit nature of
neuronal representations.
18. von Helmholtz, H. in Treatise on Physiological Optics
Vol. III 3rd edn (Voss, Hamburg, 1909).
19. MacKay, D. M. in Automata Studies (eds Shannon,
C. E. & McCarthy, J.) 235–251 (Princeton Univ. Press,
Princeton, 1956).
20. Neisser, U. Cognitive Psychology
(Appleton-Century-Crofts, New York, 1967).
21. Gregory, R. L. Perceptual illusions and brain models.
Proc. R. Soc. Lond. B Biol. Sci. 171, 179–196 (1968).
22. Gregory, R. L. Perceptions as hypotheses. Philos.
Trans. R. Soc. Lond. B Biol. Sci. 290, 181–197 (1980).
23. Ballard, D. H., Hinton, G. E. & Sejnowski, T. J. Parallel
visual computation. Nature 306, 21–26 (1983).
24. Kawato, M., Hayakawa, H. & Inui, T. A forward-inverse
optics model of reciprocal connections between visual
areas. Network: Computation in Neural Systems 4,
415–422 (1993).
25. Dayan, P., Hinton, G. E. & Neal, R. M. The Helmholtz
machine. Neural Comput. 7, 889–904 (1995).
This paper introduces the central role of generative
models and variational approaches to hierarchical
self-supervised learning and relates this to the
function of bottom-up and top-down cortical
processing pathways.
26. Lee, T. S. & Mumford, D. Hierarchical Bayesian
inference in the visual cortex. J. Opt. Soc. Am. A Opt.
Image Sci. Vis. 20, 1434–1448 (2003).
27. Kersten, D., Mamassian, P. & Yuille, A. Object
perception as Bayesian inference. Annu. Rev. Psychol.
55, 271–304 (2004).
28. Friston, K. J. A theory of cortical responses. Philos.
Trans. R. Soc. Lond. B Biol. Sci. 360, 815–836
(2005).
29. Beal, M. J. Variational Algorithms for Approximate
Bayesian Inference. Thesis, University College London
(2003).
30. Efron, B. & Morris, C. Stein’s estimation rule and its
competitors – an empirical Bayes approach. J. Am.
Stats. Assoc. 68, 117–130 (1973).
31. Kass, R. E. & Steffey, D. Approximate Bayesian
inference in conditionally independent hierarchical
models (parametric empirical Bayes models). J. Am.
Stat. Assoc. 407, 717–726 (1989).
32. Zeki, S. & Shipp, S. The functional logic of cortical
connections. Nature 335, 311–317 (1988).
Describes the functional architecture of cortical
hierarchies with a focus on patterns of anatomical
connections in the visual cortex. It emphasizes the
role of functional segregation and integration (that
is, message passing among cortical areas).
33. Felleman, D. J. & Van Essen, D. C. Distributed
hierarchical processing in the primate cerebral cortex.
Cereb. Cortex 1, 1–47 (1991).
34. Mesulam, M. M. From sensation to cognition. Brain
121, 1013–1052 (1998).
35. Sanger, T. Probability density estimation for the
interpretation of neural population codes.
J. Neurophysiol. 76, 2790–2793 (1996).
36. Zemel, R., Dayan, P. & Pouget, A. Probabilistic
interpretation of population code. Neural Comput. 10,
403–430 (1998).
37. Paulin, M. G. Evolution of the cerebellum as a
neuronal machine for Bayesian state estimation.
J. Neural Eng. 2, S219–S234 (2005).
38. Ma, W. J., Beck, J. M., Latham, P. E. & Pouget, A.
Bayesian inference with probabilistic population
codes. Nature Neurosci. 9, 1432–1438 (2006).
39. Friston, K., Mattout, J., Trujillo-Barreto, N.,
Ashburner, J. & Penny, W. Variational free energy and
the Laplace approximation. Neuroimage 34,
220–234 (2007).
40. Rao, R. P. & Ballard, D. H. Predictive coding in the
visual cortex: a functional interpretation of some
extra-classical receptive field effects. Nature Neurosci.
2, 79–87 (1998).
Applies predictive coding to cortical processing to
provide a compelling account of extra-classical
receptive fields in the visual system. It emphasizes
the importance of top-down projections in
providing predictions, by modelling perceptual
inference.
41. Mumford, D. On the computational architecture of the
neocortex. II. The role of cortico-cortical loops. Biol.
Cybern. 66, 241–251 (1992).
42. Friston, K. Hierarchical models in the brain. PLoS
Comput. Biol. 4, e1000211 (2008).
43. Murray, S. O., Kersten, D., Olshausen, B. A., Schrater, P.
& Woods, D. L. Shape perception reduces activity in
human primary visual cortex. Proc. Natl Acad. Sci.
USA 99, 15164–15169 (2002).
44. Garrido, M. I., Kilner, J. M., Kiebel, S. J. & Friston,
K. J. Dynamic causal modeling of the response to
frequency deviants. J. Neurophysiol. 101,
2620–2631 (2009).
45. Sherman, S. M. & Guillery, R. W. On the actions that
one nerve cell can have on another: distinguishing
“drivers” from “modulators”. Proc. Natl Acad. Sci. USA
95, 7121–7126 (1998).
46. Angelucci, A. & Bressloff, P. C. Contribution of
feedforward, lateral and feedback connections to the
classical receptive field center and extra-classical
receptive field surround of primate V1 neurons.
Prog. Brain Res. 154, 93–120 (2006).
47. Grossberg, S. Towards a unified theory of neocortex:
laminar cortical circuits for vision and cognition.
Prog. Brain Res. 165, 79–104 (2007).
48. Grossberg, S. & Versace, M. Spikes, synchrony, and
attentive learning by laminar thalamocortical circuits.
Brain Res. 1218, 278–312 (2008).
49. Barlow, H. in Sensory Communication (ed. Rosenblith, W.)
217–234 (MIT Press, Cambridge, Massachusetts,
1961).
50. Linsker, R. Perceptual neural organisation: some
approaches based on network models and
information theory. Annu. Rev. Neurosci. 13,
257–281 (1990).
51. Oja, E. Neural networks, principal components, and
subspaces. Int. J. Neural Syst. 1, 61–68 (1989).
52. Bell, A. J. & Sejnowski, T. J. An information
maximisation approach to blind separation and blind
de-convolution. Neural Comput. 7, 1129–1159
(1995).
53. Atick, J. J. & Redlich, A. N. What does the retina know
about natural scenes? Neural Comput. 4, 196–210
(1992).
54. Optican, L. & Richmond, B. J. Temporal encoding of
two-dimensional patterns by single units in primate
inferior cortex. III Information theoretic analysis.
J. Neurophysiol. 57, 132–146 (1987).
55. Olshausen, B. A. & Field, D. J. Emergence of simple-
cell receptive field properties by learning a sparse
code for natural images. Nature 381, 607–609
(1996).
56. Simoncelli, E. P. & Olshausen, B. A. Natural image
statistics and neural representation. Annu. Rev.
Neurosci. 24, 1193–1216 (2001).
A nice review of information theory in visual
processing. It covers natural scene statistics and
empirical tests of the efficient coding hypothesis in
individual neurons and populations of neurons.
57. Friston, K. J. The labile brain. III. Transients and
spatio-temporal receptive fields. Philos. Trans. R. Soc.
Lond. B Biol. Sci. 355, 253–265 (2000).
58. Bialek, W., Nemenman, I. & Tishby, N. Predictability,
complexity, and learning. Neural Comput. 13,
2409–2463 (2001).
59. Lewen, G. D., Bialek, W. & de Ruyter van Steveninck,
R. R. Neural coding of naturalistic motion stimuli.
Network 12, 317–329 (2001).
60. Laughlin, S. B. Efficiency and complexity in neural
coding. Novartis Found. Symp. 239, 177–187
(2001).
61. Tipping, M. E. Sparse Bayesian learning and the
Relevance Vector Machine. J. Machine Learn. Res. 1,
211–244 (2001).
62. Paus, T., Keshavan, M. & Giedd, J. N. Why do many
psychiatric disorders emerge during adolescence?
Nature Rev. Neurosci. 9, 947–957 (2008).
63. Gilestro, G. F., Tononi, G. & Cirelli, C. Widespread
changes in synaptic markers as a function of sleep and
wakefulness in Drosophila. Science 324, 109–112
(2009).
64. Roweis, S. & Ghahramani, Z. A unifying review of
linear Gaussian models. Neural Comput. 11, 305–345
(1999).
65. Hebb, D. O. The Organization of Behaviour (Wiley,
New York, 1949).
66. Paulsen, O. & Sejnowski, T. J. Natural patterns of
activity and long-term synaptic plasticity. Curr. Opin.
Neurobiol. 10, 172–179 (2000).
67. von der Malsburg, C. The Correlation Theory of Brain
Function. Internal Report 81–82, Dept. Neurobiology,
Max-Planck-Institute for Biophysical Chemistry
(1981).
68. Singer, W. & Gray, C. M. Visual feature integration and
the temporal correlation hypothesis. Annu. Rev.
Neurosci. 18, 555–586 (1995).
69. Bienenstock, E. L., Cooper, L. N. & Munro, P. W.
Theory for the development of neuron selectivity:
orientation specificity and binocular interaction in
visual cortex. J. Neurosci. 2, 32–48 (1982).
70. Abraham, W. C. & Bear, M. F. Metaplasticity: the
plasticity of synaptic plasticity. Trends Neurosci. 19,
126–130 (1996).
71. Pareti, G. & De Palma, A. Does the brain oscillate?
The dispute on neuronal synchronization. Neurol. Sci.
25, 41–47 (2004).
72. Leutgeb, S., Leutgeb, J. K., Moser, M. B. & Moser, E. I.
Place cells, spatial maps and the population code for
memory. Curr. Opin. Neurobiol. 15, 738–746
(2005).
73. Durstewitz, D. & Seamans, J. K. Beyond bistability:
biophysics and temporal dynamics of working memory.
Neuroscience 139, 119–133 (2006).
74. Anishchenko, A. & Treves, A. Autoassociative memory
retrieval and spontaneous activity bumps in small-
world networks of integrate-and-fire neurons.
J. Physiol. Paris 100, 225–236 (2006).
75. Abbott, L. F., Varela, J. A., Sen, K. & Nelson, S. B.
Synaptic depression and cortical gain control. Science
275, 220–224 (1997).
76. Yu, A. J. & Dayan, P. Uncertainty, neuromodulation
and attention. Neuron 46, 681–692 (2005).
77. Doya, K. Metalearning and neuromodulation. Neural
Netw. 15, 495–506 (2002).
78. Chawla, D., Lumer, E. D. & Friston, K. J. The
relationship between synchronization among neuronal
populations and their mean activity levels. Neural
Comput. 11, 1389–1411 (1999).
79. Fries, P., Womelsdorf, T., Oostenveld, R. & Desimone, R.
The effects of visual stimulation and selective visual
attention on rhythmic neuronal synchronization in
macaque area V4. J. Neurosci. 28, 4823–4835
(2008).
80. Womelsdorf, T. & Fries, P. Neuronal coherence during
selective attentional processing and sensory-motor
integration. J. Physiol. Paris 100, 182–193 (2006).
81. Desimone, R. Neural mechanisms for visual memory
and their role in attention. Proc. Natl Acad. Sci. USA
93, 13494–13499 (1996).
A nice review of mnemonic effects (such as
repetition suppression) on neuronal responses and
how they bias the competitive interactions between
stimulus representations in the cortex. It provides
a good perspective on attentional mechanisms in
the visual system that is empirically grounded.
82. Treisman, A. Feature binding, attention and object
perception. Philos. Trans. R. Soc. Lond. B Biol. Sci.
353, 1295–1306 (1998).
83. Maunsell, J. H. & Treue, S. Feature-based attention in
visual cortex. Trends Neurosci. 29, 317–322 (2006).
84. Spratling, M. W. Predictive-coding as a model of
biased competition in visual attention. Vision Res. 48,
1391–1408 (2008).
85. Reynolds, J. H. & Heeger, D. J. The normalization
model of attention. Neuron 61, 168–185 (2009).
86. Schroeder, C. E., Mehta, A. D. & Foxe, J. J.
Determinants and mechanisms of attentional
modulation of neural processing. Front. Biosci. 6,
D672–D684 (2001).
REVIEWS
NATURE R EVIEWS
|
NEUROSCIENCE VOLUME 11
|
FEBRUARY 2010
|
137
© 20 Macmillan Publishers Limited. All rights reserved10
87. Hirayama, J., Yoshimoto, J. & Ishii, S. Bayesian
representation learning in the cortex regulated by
acetylcholine. Neural Netw. 17, 1391–1400 (2004).
88. Edelman, G. M. Neural Darwinism: selection and
reentrant signaling in higher brain function. Neuron
10, 115–125 (1993).
89. Knobloch, F. Altruism and the hypothesis of meta-
selection in human evolution. J. Am. Acad.
Psychoanal. 29, 339–354 (2001).
90. Friston, K. J., Tononi, G., Reeke, G. N. Jr, Sporns, O. &
Edelman, G. M. Value-dependent selection in the
brain: simulation in a synthetic neural model.
Neuroscience 59, 229–243 (1994).
91. Sutton, R. S. & Barto, A. G. Toward a modern theory of
adaptive networks: expectation and prediction.
Psychol. Rev. 88, 135–170 (1981).
92. Montague, P. R., Dayan, P., Person, C. & Sejnowski,
T. J. Bee foraging in uncertain environments using
predictive Hebbian learning. Nature 377, 725–728
(1995).
A computational treatment of behaviour that
combines ideas from optimal control theory and
dynamic programming with the neurobiology of
reward. This provided an early example of value
learning in the brain.
93. Schultz, W. Predictive reward signal of dopamine
neurons. J. Neurophysiol. 80, 1–27 (1998).
94. Daw, N. D. & Doya, K. The computational
neurobiology of learning and reward. Curr. Opin.
Neurobiol. 16, 199–204 (2006).
95. Redgrave, P. & Gurney, K. The short-latency dopamine
signal: a role in discovering novel actions? Nature Rev.
Neurosci. 7, 967–975 (2006).
96. Berridge, K. C. The debate over dopamine’s role in
reward: the case for incentive salience.
Psychopharmacology (Berl.) 191, 391–431 (2007).
97. Sella, G. & Hirsh, A. E. The application of statistical
physics to evolutionary biology. Proc. Natl Acad. Sci.
USA 102 , 9541–9546 (2005).
98. Rescorla, R. A. & Wagner, A. R. in Classical
Conditioning II: Current Research and Theory (eds
Black, A. H. & Prokasy, W. F.) 64–99 (Appleton
Century Crofts, New York, 1972).
99. Bellman, R. On the Theory of Dynamic Programming.
Proc. Natl Acad. Sci. USA 38, 716–719 (1952).
100. Watkins, C. J. C. H. & Dayan, P. Q-learning. Mach.
Learn. 8, 279–292 (1992).
101. Todorov, E. in Advances in Neural Information
Processing Systems (eds Scholkopf, B., Platt, J. &
Hofmann T.) 19, 1369–1376 (MIT Press, 2006).
102. Camerer, C. F. Behavioural studies of strategic thinking
in games. Trends Cogn. Sci. 7, 225–231 (2003).
103. Smith, J. M. & Price, G. R. The logic of animal conflict.
Nature 246, 15–18 (1973).
104. Nash, J. Equilibrium points in n-person games.
Proc. Natl Acad. Sci. USA 36, 48–49 (1950).
105. Wolpert, D. M. & Miall, R. C. Forward models for
physiological motor control. Neural Netw. 9,
1265–1279 (1996).
106. Todorov, E. & Jordan, M. I. Smoothness maximization
along a predefined path accurately predicts the speed
profiles of complex arm movements. J. Neurophysiol.
80, 696–714 (1998).
107. Tseng, Y. W., Diedrichsen, J., Krakauer, J. W.,
Shadmehr, R. & Bastian, A. J. Sensory prediction-
errors drive cerebellum-dependent adaptation of
reaching. J. Neurophysiol. 98, 54–62 (2007).
108. Bays, P. M. & Wolpert, D. M. Computational
principles of sensorimotor control that minimize
uncertainty and variability. J. Physiol. 578, 387–396
(2007).
A nice overview of computational principles in
motor control. Its focus is on representing
uncertainty and optimal estimation when
extracting the sensory information required for
motor planning.
109. Shadmehr, R. & Krakauer, J. W. A computational
neuroanatomy for motor control. Exp. Brain Res. 185,
359–381 (2008).
110. Verschure, P. F., Voegtlin, T. & Douglas, R. J.
Environmentally mediated synergy between
perception and behaviour in mobile robots. Nature
425, 620–624 (2003).
111. Cohen, J. D., McClure, S. M. & Yu, A. J. Should I stay
or should I go? How the human brain manages the
trade-off between exploitation and exploration. Philos.
Trans. R. Soc. Lond. B Biol. Sci. 362, 933–942
(2007).
112. Ishii, S., Yoshida, W. & Yoshimoto, J. Control of
exploitation-exploration meta-parameter in
reinforcement learning. Neural Netw. 15, 665–687
(2002).
113. Usher, M., Cohen, J. D., Servan-Schreiber, D.,
Rajkowski, J. & Aston-Jones, G. The role of locus
coeruleus in the regulation of cognitive performance.
Science 283, 549–554 (1999).
114. Voigt, C. A., Kauffman, S. & Wang, Z. G. Rational
evolutionary design: the theory of in vitro protein
evolution. Adv. Protein Chem. 55, 79–160 (2000).
115. Freeman, W. J. Characterization of state transitions in
spatially distributed, chaotic, nonlinear, dynamical
systems in cerebral cortex. Integr. Physiol. Behav. Sci.
29, 294–306 (1994).
116. Tsuda, I. Toward an interpretation of dynamic neural
activity in terms of chaotic dynamical systems. Behav.
Brain Sci. 24, 793–810 (2001).
117. Jirsa, V. K., Friedrich, R., Haken, H. & Kelso, J. A.
A theoretical model of phase transitions in the human
brain. Biol. Cybern. 71, 27–35 (1994).
This paper develops a theoretical model (based on
synergetics and nonlinear oscillator theory) that
reproduces observed dynamics and suggests a
formulation of biophysical coupling among brain
systems.
118. Breakspear, M. & Stam, C. J. Dynamics of a
neural system with a multiscale architecture. Philos.
Trans. R. Soc. Lond. B Biol. Sci. 360, 1051–1074
(2005).
119. Bressler, S. L. & Tognoli, E. Operational principles of
neurocognitive networks. Int. J. Psychophysiol. 60,
139–148 (2006).
120. Werner, G. Brain dynamics across levels of
organization. J. Physiol. Paris 101, 273–279 (2007).
121. Pasquale, V., Massobrio, P., Bologna, L. L.,
Chiappalone, M. & Martinoia, S. Self-organization and
neuronal avalanches in networks of dissociated cortical
neurons. Neuroscience 153, 1354–1369 (2008).
122. Kitzbichler, M. G., Smith, M. L., Christensen, S. R. &
Bullmore, E. Broadband criticality of human brain
network synchronization. PLoS Comput. Biol. 5,
e1000314 (2009).
123. Rabinovich, M., Huerta, R. & Laurent, G. Transient
dynamics for neural processing. Science 321 48–50
(2008).
124. Tschacher, W. & Hake, H. Intentionality in non-
equilibrium systems? The functional aspects of self-
organised pattern formation. New Ideas Psychol. 25,
1–15 (2007).
125. Maturana, H. R. & Varela, F. De máquinas y seres
vivos (Editorial Universitaria, Santiago, 1972).
English translation available in Maturana, H. R. &
Varela, F. in Autopoiesis and Cognition (Reidel,
Dordrecht, 1980).
126. Fiorillo, C. D., Tobler, P. N. & Schultz, W. Discrete
coding of reward probability and uncertainty by
dopamine neurons. Science 299, 1898–1902
(2003).
127. Niv, Y., Duff, M. O. & Dayan, P. Dopamine,
uncertainty and TD learning. Behav. Brain Funct. 1, 6
(2005).
128. Fletcher, P. C. & Frith, C. D. Perceiving is believing: a
Bayesian approach to explaining the positive
symptoms of schizophrenia. Nature Rev. Neurosci. 10,
48–58 (2009).
129. Phillips, W. A. & Silverstein, S. M. Convergence of
biological and psychological perspectives on cognitive
coordination in schizophrenia. Behav. Brain Sci. 26,
65–82 (2003).
130. Friston, K. & Kiebel, S. Cortical circuits for perceptual
inference. Neural Netw. 22, 1093–1104 (2009).
Acknowledgments
This work was funded by the Wellcome Trust. I would like to
thank my colleagues at the Wellcome Trust Centre for
Neuroimaging, the Institute of Cognitive Neuroscience and the
Gatsby Computational Neuroscience Unit for collaborations
and discussions.
Competing interests statement
The author declares no competing financial interests.
SUPPLEMENTARY INFORMATION
See online article: S1 (box) | S2 (box) | S3 (box) | S4 (box) |
S5 (box)
ALL LINKS ARE ACTIVE IN THE ONLINE PDF
REVIEWS
138
|
FEBRUARY 2010
|
VOLUME 11 www.nature.com/reviews/neuro
© 20 Macmillan Publishers Limited. All rights reserved10
... The free energy principle proposed in Friston (2005) serves as a good starting point for an answer. It is sometimes regarded as a "unified theory of the brain" (Friston, 2010) because it attempts to explain a variety of brain processes such as perception, learning, and goal-directed action selection, based on a single objective: to minimize free energy. Free energy constitutes an upper bound on surprise, which results from interactions with the environment. ...
... this case, we can regard entropy as the expected surprise about encountered sensory states given the model (Friston, 2010). In other words: Living things must minimize expected surprises. ...
Article
Full-text available
Flexible, goal-directed behavior is a fundamental aspect of human life. Based on the free energy minimization principle, the theory of active inference formalizes the generation of such behavior from a computational neuroscience perspective. Based on the theory, we introduce an output-probabilistic, temporally predictive, modular artificial neural network architecture, which processes sensorimotor information, infers behavior-relevant aspects of its world, and invokes highly flexible, goal-directed behavior. We show that our architecture, which is trained end-to-end to minimize an approximation of free energy, develops latent states that can be interpreted as affordance maps. That is, the emerging latent states signal which actions lead to which effects dependent on the local context. In combination with active inference, we show that flexible, goal-directed behavior can be invoked, incorporating the emerging affordance maps. As a result, our simulated agent flexibly steers through continuous spaces, avoids collisions with obstacles, and prefers pathways that lead to the goal with high certainty. Additionally, we show that the learned agent is highly suitable for zero-shot generalization across environments: After training the agent in a handful of fixed environments with obstacles and other terrains affecting its behavior, it performs similarly well in procedurally generated environments containing different amounts of obstacles and terrains of various sizes at different locations.
... During prediction updating, finally, actual inputs (i.e., S2s) must be compared with predictions in order to adjust the predictive models according to the actual evidence (Friston, 2010;Seth and Friston, 2016;Barrett, 2017). Thus, it seems crucial to reappraise the S2 to diminish its affective impact, and to maximize the processing and encoding of the information, it carries about the match/mismatch with ongoing predictions. ...
Article
Full-text available
Emotion regulation (ER) strategies can influence how affective predictions are constructed by the brain (generation stage) to prearrange action (implementation stage) and update internal models according to incoming stimuli (updating stage). However, neurocomputational mechanisms by which this is achieved are unclear. We investigated through high-density EEG if different ER strategies (expressive suppression vs. cognitive reappraisal) predicted event-related potentials (ERPs) and brain source activity across affective prediction stages, as a function of contextual uncertainty. An S1-S2 paradigm with emotional faces and pictures as S1s and S2s was presented to 36 undergraduates. Contextual uncertainty was manipulated across three blocks with 100, 75, or 50% S1-S2 affective congruency. The effects of ER strategies, as assessed through the Emotion Regulation Questionnaire, on ERP and brain source activity were tested for each prediction stage through linear mixed-effects models. No ER strategy affected prediction generation. During implementation, in the 75% block, a higher tendency to suppress emotions predicted higher activity in the left supplementary motor area at 1,500–2,000 ms post-stimulus, and smaller amplitude of the Contingent Negative Variation at 2,000–2,500 ms. During updating, in the 75% block, a higher tendency to cognitively reappraise emotions predicted larger P2, Late Positive Potential, and right orbitofrontal cortex activity. These results suggest that both ER strategies interact with the levels of contextual uncertainty by differently modulating ERPs and source activity, and that different strategies are deployed in a moderately predictive context, supporting the efficient updating of affective predictive models only in the context in which model updating occurs.
... A second explanation for synchrony is the brain optimization principle (Koban et al., 2019), which holds that performing or perceiving synchronous actions requires minimal neural energy compared to performing or perceiving asynchronous actions, thus creating an economically optimal neural state. Optimization of brain functionality is based on the free energy principle, which refers to the brain's tendency to minimize coding costs when predicting and representing environmental stimuli (Friston, 2010). Neural networks have been compared to man-made electronic devices in that they are constructed to facilitate minimization of energy cost (Laughlin & Sejnowski, 2003). ...
Article
Full-text available
Interpersonal synchrony serves as a subtle, yet powerful bonding mechanism in social interactions. Problematically, the term ‘synchrony’ has been used to label a variety of distinct aspects of interpersonal coordination, such as postural similarities or movement activity entrainment. Accordingly, different algorithms have been suggested to quantify interpersonal synchrony. Yet, it remains unknown whether the different measures of synchrony represent correlated features of the same perceivable core phenomenon. The current study addresses this by comparing the suitability of a set of algorithms with respect to their association with observers’ judgments of dyadic synchrony and leader-followership. One-hundred fifteen observers viewed computer animations of characters portraying the movements of real dyads who performed a repetitive motor task with instruction to move in unison. Animations were based on full-body motion capture data synchronously collected for both partners during the joint exercise. Results showed most synchrony measures significantly correlated with (a) perceived synchrony and (b) the perceived level of balance of leading/following by each dyad member. Phase synchrony and Pearson correlations were associated most strongly with the observer ratings. This might be typical for intentional, structured forms synchrony such as ritualized group activities. It remains open if these findings also apply to spontaneous forms of synchrony as, for instance, occurring in free-running conversations.
... Next, we face-validate our formulation within a robotics context using numerical experiments (Section 4). The robotics implementation instantiates a free energy principle (FEP) approach to information processing (Friston, 2010). This allows us to modulate the (appropriate) precision parameters to solve relevant robotics challenges in perception and control; namely, state-estimation (Section 4.2.2), ...
Article
Full-text available
Computational models of visual attention in artificial intelligence and robotics have been inspired by the concept of a saliency map. These models account for the mutual information between the (current) visual information and its estimated causes. However, they fail to consider the circular causality between perception and action. In other words, they do not consider where to sample next, given current beliefs. Here, we reclaim salience as an active inference process that relies on two basic principles: uncertainty minimization and rhythmic scheduling. For this, we make a distinction between attention and salience. Briefly, we associate attention with precision control, i.e., the confidence with which beliefs can be updated given sampled sensory data, and salience with uncertainty minimization that underwrites the selection of future sensory data. Using this, we propose a new account of attention based on rhythmic precision-modulation and discuss its potential in robotics, providing numerical experiments that showcase its advantages for state and noise estimation, system identification and action selection for informative path planning.
Article
Full-text available
In the last decade, deep supervised learning has had tremendous success. However, its flaws, such as its dependency on manual and costly annotations on large datasets and being exposed to attacks, have prompted researchers to look for alternative models. Incorporating contrastive learning (CL) for self-supervised learning (SSL) has turned out as an effective alternative. In this paper, a comprehensive review of CL methodology in terms of its approaches, encoding techniques and loss functions is provided. It discusses the applications of CL in various domains like Natural Language Processing (NLP), Computer Vision, speech and text recognition and prediction. The paper presents an overview and background about SSL for understanding the introductory ideas and concepts. A comparative study for all the works that use CL methods for various downstream tasks in each domain is performed. Finally, it discusses the limitations of current methods, as well as the need for additional techniques and future directions in order to make meaningful progress in this area.
Article
Full-text available
The right posterolateral portions of the cerebellum (crus-I/II) are involved in language processing. However, their functional role in language remains unknown. The cerebellum is hypothesized to acquire an internal model that is a functional copy of mental representations in the cerebrum and to contribute to cognitive function. In this research, based on the cerebellar internal model hypothesis, we conducted task-based and resting-state functional magnetic resonance imaging (fMRI) experiments to investigate the role of the cerebellum in the syntactic and semantic aspects of comprehension of sentences. In a syntactic task, participants read sentences with center-embedded hierarchical structures. The hierarchical level-dependent activity was found in the right crus-I as well as Broca’s area (p < 0.05, voxel-based small volume correction (SVC)). In a semantic task, the participants read three types of sentences for investigation of sentence-level, phrase-level, and word-level semantic processing. The semantic level-dependent activity was found in the right crus-II as well as in the left anterior temporal lobe and the left angular gyrus (p < 0.05, voxel-based SVC). Moreover, the right crus-I/II showed significant activity when the cognitive load was high. Resting-state fMRI demonstrated intrinsic functional connectivity between the right crus-I/II and language-related regions in the left cerebrum (p < 0.05, voxel-based SVC). These findings suggest that the right crus-I and crus-II are involved, respectively, in the syntactic and semantic aspects of sentence processing. The cerebellum assists processing of language in the cerebrum when the cognitive load is high.
Chapter
Neuropsychological and experimental brain research have provided independent lines of evidence in support of cerebellar involvement in disorders of emotion. Medial cerebellar structures and their connections to the limbic system are involved in visceral aspects and the generation of emotions, whereas the posterolateral cerebello-thalamo-cortical loops are implicated in emotion regulation and subjective sense of control. Disturbances within these cerebellar-centred circuits are proposed to underlie homeostatic dysregulation and suboptimal predictive coding that provide a transdiagnostic mechanism by which the cerebellum may contribute to the vulnerability and persistence of mental disorders.
Article
Full-text available
The possibility of predicting the specific features of forthcoming environmental events is fundamental for our survival since it allows us to proactively regulate our behaviour, enhancing our chance of survival. This is particularly crucial for stimuli providing socially relevant information for communication and interaction, such as faces. While it has been consistently demonstrated that the human brain shows preferential and ontogenetically early face-evoked activity, it is unknown whether specialized neural routes are engaged by face-predictive activity early in life. In this study, we recorded high-density electrophysiological (ERP) activity in adults and 9- and 4-month-old infants undergoing an audio-visual paradigm purposely designed to predict the appearance of faces or objects starting from congruent auditory cues (i.e., human voice vs nonhuman sounds). Contingent negative variation or CNV was measured to investigate anticipatory activity as a reliable marker of stimulus expectancy even in the absence of explicit motor demand. The results suggest that CNV can also be reliably elicited in the youngest group of 4-month-old infants, providing further evidence that expectation-related anticipatory activity is an intrinsic, early property of the human cortex. Crucially, the findings also indicate that the predictive information provided by the cue (i.e., human voice vs nonhuman sounds) turns into the recruitment of different anticipatory neural dynamics for faces and objects.
Article
Full-text available
The free energy principle is notoriously difficult to understand. In this paper, we relate the principle to a framework that philosophers of biology are familiar with: Ruth Millikan’s teleosemantics. We argue that: (i) systems that minimise free energy are systems with a proper function; and (ii) Karl Friston’s notion of implicit modelling can be understood in terms of Millikan’s notion of mapping relations. Our analysis reveals some surprising formal similarities between the two frameworks, and suggests interesting lines of future research. We hope this will aid further philosophical evaluation of the free energy principle.
Article
Full-text available
Emotional experiences are proposed to arise from contextualized perception of bodily responses, also referred to as interoceptive inferences. The recognition of emotions benefits from adequate access to one's own interoceptive information. However, direct empirical evidence of interoceptive inferences and their neural basis is still lacking. In the present fMRI study healthy volunteers performed a probabilistic emotion classification task with videotaped dynamically unfolding facial expressions. In a first step, we aimed to determine functional areas involved in the processing of dynamically unfolding emotional expressions. We then tested whether individuals with higher interoceptive accuracy (IAcc), as assessed by the Heartbeat detection task (HDT), or higher interoceptive sensitivity (IS), as assessed by the Multidimensional Assessment of Interoceptive Awareness, Version 2 (MAIA-2), benefit more from the contextually given likelihood of emotional valence and whether brain regions reflecting individual IAcc and/or IS play a role in this. Individuals with higher IS benefitted more from the biased probability of emotional valence. Brain responses to more predictable emotions elicited a bilateral activity pattern comprising the inferior frontal gyrus and the posterior insula. Importantly, individual IAcc scores positively covaried with brain responses to more surprising and less predictable emotional expressions in the insula and caudate nucleus. We show for the first time that IAcc score is associated with enhanced processing of interoceptive prediction errors, particularly in the anterior insula. A higher IS score seems more likely to be associated with a stronger weighting of attention to interoceptive changes processed by the posterior insula and ventral prefrontal cortex.
Article
Full-text available
A computational view of how perception and cognition can be modeled as dynamic patterns of transient activity within neural networks.
Book
Stuart Kauffman here presents a brilliant new paradigm for evolutionary biology, one that extends the basic concepts of Darwinian evolution to accommodate recent findings and perspectives from the fields of biology, physics, chemistry and mathematics. The book drives to the heart of the exciting debate on the origins of life and maintenance of order in complex biological systems. It focuses on the concept of self-organization: the spontaneous emergence of order widely observed throughout nature. Kauffman here argues that self-organization plays an important role in the emergence of life itself and may play as fundamental a role in shaping life's subsequent evolution as does the Darwinian process of natural selection. Yet until now no systematic effort has been made to incorporate the concept of self-organization into evolutionary theory. The construction requirements which permit complex systems to adapt remain poorly understood, as is the extent to which selection itself can yield systems able to adapt more successfully. This book explores these themes. It shows how complex systems, contrary to expectations, can spontaneously exhibit stunning degrees of order, and how this order, in turn, is essential for understanding the emergence and development of life on Earth. Topics include the new biotechnology of applied molecular evolution, with its important implications for developing new drugs and vaccines; the balance between order and chaos observed in many naturally occurring systems; new insights concerning the predictive power of statistical mechanics in biology; and other major issues. Indeed, the approaches investigated here may prove to be the new center around which biological science itself will evolve. The work is written for all those interested in the cutting edge of research in the life sciences.
Article
In recent years, many new cortical areas have been identified in the macaque monkey. The number of identified connections between areas has increased even more dramatically. We report here on (1) a summary of the layout of cortical areas associated with vision and with other modalities, (2) a computerized database for storing and representing large amounts of information on connectivity patterns, and (3) the application of these data to the analysis of hierarchical organization of the cerebral cortex. Our analysis concentrates on the visual system, which includes 25 neocortical areas that are predominantly or exclusively visual in function, plus an additional 7 areas that we regard as visual-association areas on the basis of their extensive visual inputs. A total of 305 connections among these 32 visual and visual-association areas have been reported. This represents 31% of the possible number of pathways it each area were connected with all others. The actual degree of connectivity is likely to be closer to 40%. The great majority of pathways involve reciprocal connections between areas. There are also extensive connections with cortical areas outside the visual system proper, including the somatosensory cortex, as well as neocortical, transitional, and archicortical regions in the temporal and frontal lobes. In the somatosensory/motor system, there are 62 identified pathways linking 13 cortical areas, suggesting an overall connectivity of about 40%. Based on the laminar patterns of connections between areas, we propose a hierarchy of visual areas and of somato sensory/motor areas that is more comprehensive than those suggested in other recent studies. The current version of the visual hierarchy includes 10 levels of cortical processing. Altogether, it contains 14 levels if one includes the retina and lateral geniculate nucleus at the bottom as well as the entorhinal cortex and hippocampus at the top. Within this hierarchy, there are multiple, intertwined processing streams, which, at a low level, are related to the compartmental organization of areas V1 and V2 and, at a high level, are related to the distinction between processing centers in the temporal and parietal lobes. However, there are some pathways and relationships (about 10% of the total) whose descriptions do not fit cleanly into this hierarchical scheme for one reason or another. In most instances, though, it is unclear whether these represent genuine exceptions to a strict hierarchy rather than inaccuracies or uncertainties in the reported assignment.
Article
Introduction: Debate continues over the precise causal contribution made by mesolimbic dopamine systems to reward. There are three competing explanatory categories: 'liking', learning, and 'wanting'. Does dopamine mostly mediate the hedonic impact of reward ('liking')? Does it instead mediate learned predictions of future reward, prediction error teaching signals and stamp in associative links (learning)? Or does dopamine motivate the pursuit of rewards by attributing incentive salience to reward-related stimuli ('wanting')? Each hypothesis is evaluated here, and it is suggested that the incentive salience or 'wanting' hypothesis of dopamine function may be consistent with more evidence than either learning or 'liking'. In brief, recent evidence indicates that dopamine is neither necessary nor sufficient to mediate changes in hedonic 'liking' for sensory pleasures. Other recent evidence indicates that dopamine is not needed for new learning, and not sufficient to directly mediate learning by causing teaching or prediction signals. By contrast, growing evidence indicates that dopamine does contribute causally to incentive salience. Dopamine appears necessary for normal 'wanting', and dopamine activation can be sufficient to enhance cue-triggered incentive salience. Drugs of abuse that promote dopamine signals short circuit and sensitize dynamic mesolimbic mechanisms that evolved to attribute incentive salience to rewards. Such drugs interact with incentive salience integrations of Pavlovian associative information with physiological state signals. That interaction sets the stage to cause compulsive 'wanting' in addiction, but also provides opportunities for experiments to disentangle 'wanting', 'liking', and learning hypotheses. Results from studies that exploited those opportunities are described here. Conclusion: In short, dopamine's contribution appears to be chiefly to cause 'wanting' for hedonic rewards, more than 'liking' or learning for those rewards.
Article
Ablation and single-unit studies in primates have shown that inferior temporal (IT) cortex is important for pattern discrimination. The first paper in this series suggested that single units in IT cortex of alert monkeys respond to a set of two-dimensional patterns with complex temporal modulation of their spike trains. The second paper quantified the waveform of the modulated responses of IT neurons with principal components and demonstrated that the coefficients of two to four of the principal components were stimulus dependent. Although the coefficients of the principal components are uncorrelated, it is possible that they are not statistically independent. That is, several coefficients could be determined by the same feature of the stimulus, and thus could be conveying the same information. The final part of this study examined this issue by comparing the amount of information about the stimulus that can be conveyed by two codes: a temporal waveform code derived from the coefficients of the first three principal components and a mean rate code derived from the spike count. We considered the neuron to be an information channel conveying messages about stimulus parameters. Previous applications of information theory to neurophysiology have dealt either with the theoretical capacity of neuronal channels or the temporal distribution of information within the spike train. This previous work usually used a general binary code to represent the spike train of a neuron&apos;s response. Such a general approach yields no indication of the nature of the neuron&apos;s intrinsic coding scheme because it depends only on the timing of spikes in the response. In particular, it is independent of any statistical properties of the responses. Our approach uses the principal components of the response waveform to derive a code for representing information about the stimuli. We regard this code as an indication of the neuron&apos;s intrinsic coding scheme, because it is based on the statistical properties of the neuronal responses. We measured how much information about the stimulus was present in the neuron&apos;s responses. This transmitted information was calculated for codes based on either the spike count or on the first three principal components of the response waveform. The information transmitted by each of the first three principal components was largely independent of that transmitted by the others. It was found that the average amount of information transmitted by the principal components was about twice as large as that transmitted by the spike count.(ABSTRACT TRUNCATED AT 400 WORDS)
Article
The development of stimulus selectivity in the primary sensory cortex of higher vertebrates is considered in a general mathematical framework. A synaptic evolution scheme of a new kind is proposed in which incoming patterns rather than converging afferents compete. The change in the efficacy of a given synapse depends not only on instantaneous pre- and postsynaptic activities but also on a slowly varying time-averaged value of the postsynaptic activity. Assuming an appropriate nonlinear form for this dependence, development of selectivity is obtained under quite general conditions on the sensory environment. One does not require nonlinearity of the neuron's integrative power nor does one need to assume any particular form for intracortical circuitry. This is first illustrated in simple cases, e.g., when the environment consists of only two different stimuli presented alternately in a random manner. The following formal statement then holds: the state of the system converges with probability 1 to points of maximum selectivity in the state space. We next consider the problem of early development of orientation selectivity and binocular interaction in primary visual cortex. Giving the environment an appropriate form, we obtain orientation tuning curves and ocular dominance comparable to what is observed in normally reared adult cats or monkeys. Simulations with binocular input and various types of normal or altered environments show good agreement with the relevant experimental data. Experiments are suggested that could test our theory further.