Klaus Obermayer

Technische Universität Berlin | TUB · Department of Software Engineering and Theoretical Computer Science

We adapt non-linear optimal control theory (OCT) to control oscillations and network synchrony and apply it to models of neural population dynamics. OCT is a mathematical framework to compute an efficient stimulation for dynamical systems. In its standard formulation, it requires a well-defined reference trajectory as target state. This requirement...

Even if the scene before our eyes remains static for some time, we might explore it differently compared with how we examine static images, which are commonly used in studies on visual attention. Here we show experimentally that the top-down expectation of changes in natural scenes causes clearly distinguishable gaze behavior for visually identical...

How we perceive objects around us depends on what we actively attend to, yet our eye movements depend on the perceived objects. Still, object segmentation and gaze behavior are typically treated as two independent processes. Drawing on an information processing pattern from robotics, we present a mechanistic model that simulates these processes for...

In this paper, we explore the necessity of meta-training the final layer of the network in model-agnostic meta-learning (MAML) for few-shot learning. Previous research has shown that updating only the final layer during fine-tuning can improve performance. We go beyond this by randomly re-initializing the final layer before optimizing the inner loo...

Introduction We examined changes in large-scale functional connectivity and temporal dynamics and their underlying mechanisms in schizophrenia (ScZ) through measurements of resting-state functional magnetic resonance imaging (rs-fMRI) data and computational modelling. Methods The rs-fMRI measurements from patients with chronic ScZ (n=38) and match...

Study objectives We aimed to build a tool which facilitates manual labeling of sleep slow oscillations (SOs) and evaluate the performance of traditional sleep SO detection algorithms on such a manually labeled data set. We sought to develop improved methods for SO detection. Method SOs in polysomnographic recordings acquired during nap time from t...

We examined changes in large-scale functional connectivity and temporal dynamics and their underlying mechanisms in schizophrenia (ScZ) through measurements of resting-state functional magnetic resonance imaging (rs-fMRI) data and computational modelling. The rs-fMRI measurements from patients with chronic ScZ (n=38) and matched healthy controls (n...

The complexity of natural scenes makes it challenging to experimentally study the mechanisms behind human gaze behavior when viewing dynamic environments. Historically, eye movements were believed to be driven primarily by space-based attention towards locations with salient features. Increasing evidence suggests, however, that visual attention doe...

Certain neurophysiological characteristics of sleep, in particular slow oscillations (SOs), sleep spindles, and their temporal coupling, have been well characterised and associated with human memory abilities. Delta waves, which are somewhat higher in frequency and lower in amplitude compared to SOs, and their interaction with spindles have only re...

Nonlinear dynamical systems describe neural activity at various scales and are frequently used to study brain functions and the impact of external perturbations. Here, we explore methods from optimal control theory (OCT) to study efficient, stimulating "control" signals designed to make the neural activity match desired targets. Efficiency is quant...

The complexity of natural scenes makes it challenging to experimentally study the mechanisms behind human gaze behavior when viewing dynamic environments. Historically, eye movements were believed to be driven primarily by bottom-up saliency, but increasing evidence suggests that objects also play a significant role in guiding attention. We present...

Objective: Feedback training is a practical approach to brain-computer interface (BCI) end-users learning to modulate their sensorimotor rhythms (SMR). BCI self-regulation learning has been shown to be influenced by subjective psychological factors, such as motivation. However, few studies have taken into account the users' self-motivation as addi...

Graphics processing units (GPUs) are widely available and have been used with great success to accelerate scientific computing in the last decade. These advances, however, are often not available to researchers interested in simulating spiking neural networks, but lacking the technical knowledge to write the necessary low-level code. Writing low-le...

Background: Oscillatory rhythms during sleep, such as slow oscillations (SOs) and spindles and, most importantly, their coupling, are thought to underlie processes of memory consolidation. External slow oscillatory transcranial direct current stimulation (so-tDCS) with a frequency of 0.75 Hz has been shown to improve this coupling and memory conso...

We apply the framework of nonlinear optimal control to a biophysically realistic neural mass model, which consists of two mutually coupled populations of deterministic excitatory and inhibitory neurons. External control signals are realized by time-dependent inputs to both populations. Optimality is defined by two alternative cost functions that tr...

It is common practice to reuse models initially trained on different data to increase downstream task performance. Especially in the computer vision domain, ImageNet-pretrained weights have been successfully used for various tasks. In this work, we investigate the impact of transfer learning for segmentation problems, being pixel-wise classificatio...

In transfer learning, only the last part of the networks - the so-called head - is often fine-tuned. Representation similarity analysis shows that the most significant change still occurs in the head even if all weights are updatable. However, recent results from few-shot learning have shown that representation change in the early layers, which are...

We provide a new algorithm for solving Risk Sensitive Partially Observable Markov Decisions Processes, when the risk is modeled by a utility function, and both the state space and the space of observations is finite. This algorithm is based on an observation that the change of measure and the subsequent introduction of the information space that is...

Sleep manifests itself by the spontaneous emergence of characteristic oscillatory rhythms, which often time-lock and are implicated in memory formation. Here, we analyze a neural mass model of the thalamocortical loop in which the cortical node can generate slow oscillations (approximately 1 Hz) while its thalamic component can generate fast sleep...

Gamma rhythms play a major role in many different processes in the brain, such as attention, working memory, and sensory processing. While typically considered detrimental, counterintuitively noise can sometimes have beneficial effects on communication and information transfer. Recently, Meng and Riecke showed that synchronization of interacting ne...

During slow-wave sleep, the brain is in a self-organized regime in which slow oscillations (SOs) between up- and down-states travel across the cortex. While an isolated piece of cortex can produce SOs, the brain-wide propagation of these oscillations are thought to be mediated by the long-range axonal connections. We address the mechanism of how SO...

Real-time load information in public transport is of high importance for both passengers and service providers. Neural algorithms have shown a high performance on various object counting tasks and play a continually growing methodological role in developing automated passenger counting systems. However, the publication of public-space video footage...

Gamma rhythms play a major role in many different processes in the brain, such as attention, working memory and sensory processing. While typically considered detrimental, counterintuitively noise can sometimes have beneficial effects on communication and information transfer. Recently, Meng and Riecke showed that synchronization of interacting net...

In this article, we give an interactive introduction to model-agnostic meta-learning (MAML), a well-establish method in the area of meta-learning. Meta-learning is a research field that attempts to equip conventional machine learning architectures with the power to gain meta-knowledge about a range of tasks to solve problems like the one above on a...

neurolib is a computational framework for whole-brain modeling written in Python. It provides a set of neural mass models that represent the average activity of a brain region on a mesoscopic scale. In a whole-brain network model, brain regions are connected with each other based on biologically informed structural connectivity, i.e., the connectom...

Background: Oscillatory rhythms during sleep such as slow oscillations (SO) and spindles, and most importantly their coupling, are thought to underlie processes of memory consolidation. External slow oscillatory transcranial direct current stimulation (so-tDCS) with a frequency of 0.75 Hz has been shown to improve this coupling and memory consolida...

Certain neurophysiological characteristics of sleep, in particular slow oscillations (SO), sleep spindles, and their temporal coupling, have been well characterized and associated with human memory formation. Delta waves, which are somewhat higher in frequency and lower in amplitude compared to SO, have only recently been found to play a critical r...

Sleep manifests itself by the spontaneous emergence of characteristic oscillatory rhythms, which often time-lock and are implicated in the memory formation. Here, we analyze a neural mass model of the thalamo-cortical loop of which the cortical node can generate slow oscillations (approx. 1 Hz) while its thalamic component can generate sleep spindl...

While the prediction of fixation locations on static scenes is an active research area, neither the sequencing of saccades nor dynamic stimuli have received much attention in computational modeling. Here we present a new framework for the simulation of human exploration behavior in dynamic real-world scenes. Saccades are modeled as a sequential dec...

We apply the framework of optimal nonlinear control to steer the dynamics of a whole-brain network of FitzHugh-Nagumo oscillators. Its nodes correspond to the cortical areas of an atlas-based segmentation of the human cerebral cortex, and the internode coupling strengths are derived from diffusion tensor imaging data of the connectome of the human...

The coordinated dynamic interactions of large-scale brain circuits and networks have been associated with cognitive functions and behavior. Recent advances in network neuroscience have suggested that the anatomical organization of such networks puts fundamental constraints on the dynamical landscape of brain activity, i.e., the different states, or...

Visually exploring the world around us is not a passive process. Instead, we actively explore the world and acquire visual information over time. Here, we present a new model for simulating human eye-movement behavior in dynamic real-world scenes. We model this active scene exploration as a sequential decision making process. We adapt the popular d...

Brain–computer interface (BCI) has developed rapidly over the past two decades, mainly due to advancements in machine learning. Subjects must learn to modulate their brain activities to ensure a successful BCI. Feedback training is a practical approach to this learning process; however, the commonly used classifier-dependent approaches have inheren...

[This corrects the article DOI: 10.1371/journal.pcbi.1007822.].

The coordinated, dynamical interactions of large-scale networks give rise to cognitive function. Recent advances in network neuroscience have suggested that the anatomical organization of such networks puts a fundamental constraint on the dynamical landscape of the brain. Consequently, changes in large-scale brain activity have been hypothesized to...

neurolib is a computational framework for whole-brain modeling written in Python. It provides a set of neural mass models that represent the average activity of a brain region on a mesoscopic scale. In a whole-brain network model, brain regions are connected with each other based on biologically informed structural connectivity, i.e. the connectome...

We apply the framework of optimal nonlinear control to steer the dynamics of a whole-brain network of FitzHugh-Nagumo oscillators. Its nodes correspond to the cortical areas of an atlas-based segmentation of the human cerebral cortex, and the inter-node coupling strengths are derived from Diffusion Tensor Imaging data of the connectome of the human...

During slow-wave sleep, the brain is in a self-organized regime in which slow oscillations (SOs) between up- and down-states propagate across the cortex. We address the mechanism of how SOs emerge and can recruit large parts of the brain using a whole-brain model based on empirical connectivity data. Individual brain areas generate SOs that are ind...

In visual areas of primates, neurons activate in parallel while the animal is engaged in a behavioral task. In this study, we examine the structure of the population code while the animal performs delayed match-to-sample tasks on complex natural images. The macaque monkeys visualized two consecutive stimuli that were either the same or different, w...

Electrical stimulation of neural systems is a key tool for understanding neural dynamics and ultimately for developing clinical treatments. Many applications of electrical stimulation affect large populations of neurons. However, computational models of large networks of spiking neurons are inherently hard to simulate and analyze. We evaluate a red...

Primary visual cortex (V1) is absolutely necessary for normal visual processing, but whether V1 encodes upcoming behavioral decisions based on visual information is an unresolved issue, with conflicting evidence. Further, no study so far has been able to predict choice from time-resolved spiking activity in V1. Here, we hypothesized that the choice...

Identification and localization of sounds are both integral parts of computational auditory scene analysis. Although each can be solved separately, the goal of forming coherent auditory objects and achieving a comprehensive spatial scene understanding suggests pursuing a joint solution of the two problems. This article presents an approach that rob...

We study a mean-field neural mass model of spiking adaptive exponential integrate-and-fire (AdEx) neurons and are interested in the effects of electric fields on the population activity. Our primary research question is: How do time-varying electric field inputs affect endogenous population dynamics of a cortical E-I network?

We propose a new model of the read-out of spike trains that exploits the multivariate structure of responses of neural ensembles. Assuming the point of view of a read-out neuron that receives synaptic inputs from a population of projecting neurons, synaptic inputs are weighted with a heterogeneous set of weights. We propose that synaptic weights re...

We propose a new algorithm that uses an auxiliary Neural Network to calculate the transport distance between two data distributions and export an optimal transport map. In the sequel we use the aforementioned map to train Generative Networks. Unlike WGANs, where the Euclidean distance is implicitly used, this new method allows to use any transporta...

Electrical stimulation of neural populations is a key tool for understanding neural dynamics and developing treatments. To investigate the effects of external stimulation on a basic cortical motif, we analyse the dynamical properties of an efficient mean-field neural mass model of excitatory and inhibitory adaptive exponential integrate-and-fire (A...

In visual areas of primates, neurons activate in parallel while the animal is engaged in a behavioral task. In this study, we examine the structure of the population code while the animal performs delayed match to sample task on complex natural images. The macaque monkeys visualized two consecutive stimuli that were either the same or different, wh...

We propose a novel method of the read-out of spike trains that exploits the structure of responses of neural ensembles. Assuming the point of view of a read-out neuron that receives synaptic inputs from a population of projecting neurons, synaptic inputs are weighted with a heterogeneous set of weights. We propose that synaptic weights reflect the...

By using the fact that the space of all probability measures with finite support can be completed in two different fashions, one generating the Arens-Eells space and another generating the Kantorovich-Wasserstein (Wasserstein-1) space, and by exploiting the duality relationship between the Arens-Eells space with the space of Lipschitz functions, we...

Transcranial brain stimulation and evidence of ephaptic coupling have sparked strong interests in understanding the effects of weak electric fields on the dynamics of neuronal populations. While their influence on the subthreshold membrane voltage can be biophysically well explained using spatially extended neuron models, mechanistic analyses of ne...

Subthreshold response properties of ball-and-stick neurons and fitted two-compartment neurons. A: amplitude of subthreshold somatic impedances for inputs at the soma and dendrite, respectively, as a function of input frequency (cf. Fig 1C). B: amplitude of subthreshold somatic voltage responses to a sinusoidal electric field with amplitude E1 = 1 V...

Spike rate responses of two-compartment neurons. A: spike rate as a function of somatic (green) and dendritic (blue) mean input, respectively, for σs/Cs=0.1V/s and σd/Cd=0.05V/s (cf. Fig 1H). B: amplitude of spike rate responses to sinusoidal modulations of the mean input at the soma (green) or at the dendrite (blue) as a function of modulation fre...

Identification and localization of sounds are both integral parts of computational auditory scene analysis. Although each can be solved separately, the goal of forming coherent auditory objects and achieving a comprehensive spatial scene understanding suggests pursuing a joint solution of the two problems. This work presents an approach that robust...

Computational auditory scene analysis is gaining interest in the last years. Trailing behind the more mature field of speech recognition, it is particularly general sound event detection, that is attracting increasing attention. Crucial for training and testing reasonable models is having available enough suitable data -- until recently, general so...

Classifying objects in images is a major task in many modern computer vision applications. Due to the complex­ity of detecting, describing and finally recognizing items with differing sizes and orientations, this task still remains a highly challenging problem. During the last decades, Con­volutional Neural Networks (CNNs) have proven to yield top­...

This paper is in a publication limbo at the moment. In the last review, we were crtisized for lack of examples and proper connection to other results. If someone finds something usable/relatable in the method that the paper describes, we are open to any form of collaboration. The core novelty/observetion in this paper is that the change of measure...

Transcranial brain stimulation and evidence of ephaptic coupling have sparked strong interests in understanding the effects of weak electric fields on the dynamics of neuronal populations. While their influence on the subthreshold membrane voltage can be biophysically well explained using spatially extended neuron models, mechanistic analyses of ne...

The rise of transcranial current stimulation (tCS) techniques have sparked an increasing interest in the effects of weak extracellular electric fields on neural activity. These fields modulate ongoing neural activity through polarization of the neuronal membrane. While the somatic polarization has been investigated experimentally, the frequency-dep...

Increasing the neuron membrane surface reduces the sensitivity to DC fields. Polarization of passive cells due to a positive 1 V/m field plotted as the function of the distance from the soma. For clarity basal dendrites are plotted with negative distance. The polarization is plotted for original cell (green) and for a cell with an increased membran...

In a passive cable with a right bending angle, the length of the bent branch mainly affects the field sensitivity at its extremity, without inducing a resonance. The subplots represent the field sensitivity (in V / (V/λ)) at both cable ends: (top,red star) the unbent branch and (bottom, violet circle) the bent one, as function of the field frequenc...

In a passive cable with an acute bending angle, the membrane time constant τ determines the resonance frequency. The subplots represent the field sensitivity (in V / (V/λ)) at both cable ends: (top,red star) the unbent branch and (bottom, violet circle) the bent one, as function of the field frequency (x axis). The field sensitivity are displayed f...

In an active pyramidal cell model, the field sensitivity presents a strong resonance around 10-20Hz at the apical dendrite but not at the soma or the basal dendrite. Frequency-dependent sensitivity of the cell to AC field measured at different locations at the basal (A) or apical (B) dendrites. Colors code the distance to the soma of the considered...

The distribution of the resting membrane potential does not qualitatively affects the field sensitivity profile of the active cell at the apical dendrites. (A) Distribution of the membrane potential at rest, i.e. in absence of electrical fields, in the fully active Hay et al. model. (B) Field sensitivity of the fully active model in case of non-uni...

A passive pyramidal neuron exhibits a frequency resonance in its field sensitivity at the proximal dendrites. (Left) Frequency-dependent sensitivity of the passive cell, i.e. without any active ion channels, due to AC field parallel to the somato-dendritic axis. The locations where these sensitivities are measured are displayed on the cell (right)....

In a passive cable with an obtuse bending angle, the length of the bent branch has little effects on the field sensitivity at both extremities. The subplots represent the field sensitivity (in V / (V/λ)) at both cable ends: (top,red star) the unbent branch and (bottom, violet circle) the bent one, as function of the field frequency (x axis). The fi...