Technical ReportPDF Available

MATLAB Cheat Sheet for Data Science - London School of Economics

Authors:

Abstract

* Preliminaries: Ways to get help, File extensions, Common data types, Data import/export, Basic commands, Create basic variables, Basic math functions, Trigonometric functions, Linear algebra, Accessing/assignment elements, Character and string, Regular expression, "IS*" functions, Convert functions, Programming, Errors, Parallel computing (CPU & GPU), Plotting & Figures. * Data science: Neural network, Support vector machines/regression, Deep learning, Decision tree, Linear/Nonlinear regression, Clustering, Dimension reduction/feature selection, Cross-validation.
MATLAB Cheat Sheet for Data Science
Ali Habibnia Eghbal Rahimikia
Last update: May, 2017
MATLAB r2017a
Preliminaries
fuc : This function or command requires a toolbox to execute.
Ways to get help
doc Display documentation.
doc command Display documentation for function.
help Display documentation in command window.
help command Display help text in command window.
lookfor (X) Search all M-files for X.
docsearch (X) Search documentation for X.
demo Access demonstration examples.
which command Locate functions.
File extensions
.m A MATLAB script, function, or class.
.mat A MATLAB data, stores workspace.
.fig MATLAB figure or GUI template.
.p MATLAB protected function file.
.mlx MATLAB live script.
.mex MATLAB executable.
.mlapp MATLAB App Designer template.
.mdl .slx Simulink model.
.mdlp .slxp Simulink protected model.
.mlappinstall MATLAB app installer file.
.mltbx MATLAB toolbox file.
Common data types
single Single precision numerical data (32 bits).
double Double precision numerical data (64 bits).
char Character array.
string String array.
logical True (1) or false (0).
struct Structure array.
cell Cell array.
map container Map values to unique keys (dictionary).
Data import/export
xlsread/xlswrite Read/write Excel spreadsheet.
load/save Load/save MATLAB variables.
load/save -ascii Load/save text files (.txt, .csv).
dlmread/dlmwrite Read/write ASCII-delimited file.
readtable/writetable Create/write table from file.
fscanf/fprintf Read/write data from/to text file.
textscan Read formatted data from text file.
fgetl Read line from file, removing
newline characters.
fgets Read line from file, keeping
newline characters.
fread/fwrite Read/write from/to binary file.
fopen/fclose Open/close file.
importdata Load data from file.
readall Read data from data-store.
imread/imwrite Read/write image file.
save filename Save all variables to .mat file.
save filename x,y Save x,yvariables to .mat file.
load filename Load all variables from .mat file.
webread/webwrite (URL) Read/write content from/to URL.
websave (URL) Save data from URL to file.
Basic commands
clc Clear command window.
clear Clear workspace.
clear (X) Clear (X) from memory.
close (X) Close figure (X).
close all Close all figures.
... Continue entering statement.
clf Clear current figure.
whos (X) Size, bytes, class, and attributes of (X).
ver List MATLAB version and toolboxes.
dir List current folder contents.
tic/toc Start/stop stopwatch timer.
beep Produce system beep sound.
ans Last answer.
pwd Current directory.
path View/change search directory.
pathtool Open set path window.
mkdir Make new directory.
cd Change current directory.
what List of MATLAB files in folder.
which Find directory of functions.
lasterr Last error message.
lastwarn Last warning message.
rehash Refresh caches.
home Send cursor home.
exit Close MATLAB.
Create basic variables
x=5 Define variable xto be 5.
x=nan Define variable xto be Not-a-Number.
j:k Row vector from jto k(step size: 1).
j:i:k Row vector from jto k(step size: i).
linspace(a,b,n) n numbers linearly spaced between aand b.
logspace(a,b,n) n numbers logarithmically spaced between aand b.
NaN(a,b) a ×bmatrix of NaN values.
ones(a,b) a ×bmatrix of 1 values.
zeros(a,b) a ×bmatrix of 0 values.
eye(a) Identity matrix of size a.
sparse(a,b) a ×bsparse matrix.
rand(a,b) Uniform a×brandom numbers in [0,1).
randi(imax,a,b) Uniform a×brandom integers in [1,imax].
randn(a,b) Gaussian a×brandom numbers.
randperm(a) Integer random permutation in [1,a].
diag(x) Square matrix (vector x: diagonal elements).
Basic math functions
abs(x) Absolute value of x.
sqrt(x) Square root of x.
sign(x) Sign of x.
round(x) Round of x.
ceil(x) Round xtoward positive infinity.
fix(x) Round xtoward zero.
floor(x) Round xtoward negative infinity.
complex(a,b) Complex array (z=a+bi).
real(x) Real part of complex number.
image(x) Imaginary part of complex number.
conj(x) Complex conjugate of x.
log(x) Natural logarithm of x.
log10(x) Common logarithm of x.
exp(x) Exponential of x(ex).
rem(a,b) Remainder after division of aby b.
mod(a,b) Remainder after division of aby b(modulo operation).
lcm(a,b) Least common multiples of aand b.
gcd(a,b) Greatest common multiples of aand b.
nthroot(a,n) Real n-th root of a.
Trigonometric functions
#: sin, cos, tan, sec, or cot. sine, cosine, tangent, secant, or cotangent.
#/#d(x) # of xin radians/degrees.
#h(x) Hyperbolic #of x.
a#/a#d(x) Inverse #of xin radians/degrees.
a#h(x) Inverse hyperbolic #of x.
atan2/atan2d(x) Four-quadrant inverse tan of xin
radians/degrees.
hypot(x) Square root of sum of squares of x.
deg2rad(x) Convert xfrom degrees to radians.
rad2deg(x) Convert xfrom radians to degrees.
MATLAB Cheat Sheet for Data Science - London School of Economics. 1 of 9
Linear algebra
x=[1,2,3] 1×3 vector (double array).
x=[1;2;3] 3×1 vector.
x=[1,2;3,4] 2×2 matrix (double array).
x=[1,2;3,4;5,6] 3×2 matrix.
x={1,‘st’} 1×2 cell array.
sx.x1=[1,2,3] 1×3 vector stored in sx structure array.
sx.x2={1,‘st’} 1×2 cell stored in sx structure array.
x*y Matrix multiplication.
x+y Element by element addition.
x-y Element by element subtraction.
x.*y Element by element multiplication.
x./y Element by element division.
A^n Normal matrix power of A.
A.^n Element-wise power of A.
A’ Transpose of A.
inv(A) Inverse.
size(A) Size (rows and columns).
numel(A) Number of elements.
min(A) Smallest elements.
cummin(A) Cumulative minimum of array elements.
max(A) Largest elements.
cummax(A) Cumulative maximum of array elements.
sum(A) Sum of array elements.
cumsum(A) Cumulative sum of array elements.
mean(A) Average of elements.
median(A) Median of elements.
mode(A) Mode of elements.
prod(A) Product of array elements.
cumprod(A) Cumulative product of array elements.
diff(x,k) Successive k-differences of x.
std(A) Standard deviation.
var(A) Variance.
cov(A) Covariance.
corrcoef(A) Correlation coefficients (columns: random
variables, rows: observations).
eig(A) Eigenvalues and eigenvectors.
svd(A) Singular values.
norm(A) Norms.
sort(A) Sorts vector from smallest to largest.
sortrows(A) Sorts rows of Ain ascending order.
rank(A) Rank.
chol(A) Cholesky factorization of matrix.
det(A) Determinant of square matrix.
factor(A) Prime factors.
perm(A) Permutations of the elements.
Accessing/assignment elements
If x is a vector:
x(i) Element i-th of x(i, j 1).
x(i:j) Elements from the i-th to the j-th of x.
x(i:end) Elements from the i-th to the last one of x.
x(:) All the elements of x.
x([i,j]) i-th and j-th elements.
If A is a matrix:
A(i,j) * Element i,jof A(i, j, l, m 1).
A(i) * Element i-th of A.
A(i:j,l:m) * Elements in rows from ito jwhich are
in columns from lto m.
A([a,b],[c,d]) *a-th and b-th elements in rows and c-th
and d-th elements in columns.
A(:,i) *i-th column elements.
A(:,[a,b]) a-th and b-th columns elements.
A(i,:) *i-th row elements.
A([i,j],:) *i-th and j-th rows elements.
A>i Logical array of elements higher than i.
find(A>i) Indices of elements higher than i.
find(A==i) Indices of elements equal to i.
diag(A) Elements in the principal diagonal of A.
tril(A) Lower triangular part of A.
triu(A) Upper triangular part of A.
A(i,j)=a Replace element i,jof Awith a.
A(:,i)=[a;b] Replace i-th column of Awith [a;b].
A(i,:)=[] * Delete i-th row of A.
A([i,j],:)=[] * Delete i-th and j-th rows of A.
A(A>m)=v Replace all elements over mwith v.
A(A==m)=v Replace all elements is equal to mwith v.
arrayfun(func,A) Apply a function to each element of A.
bsxfun(func,A,B) Apply an element-wise binary operation
specified by func to Aand B.
If A is a cell:
Above matrix operations which are marked
with asterisk(*). Output is cell array of
elements.
A{i,j} Element i,jof A(i, j, l, m 1).
A(i,j)={a} Replace element i,jof Awith cell {a}.
A(:,i)={a;b} Replace i-th column of Awith cell {a;b}.
cellfun(func,A) Apply a function to each cell in A.
Character and string
st: string.
char: character.
x=‘text’ Define variable xto be ‘text’ (char).
x(i) i-th part of char.
x(i:j) i-th to j-th part of char.
x=string(‘text’) Define variable xto be ‘text’ (str).
x={‘s1’,‘s2’} Define more than one char.
x=[string(‘st1’)...
,string(‘st2’)] Define more than one str.
x{i,j} Element i,j of x(char).
x(i,j) Element i,j of x(str).
x{i}(j) i-th part of j-th char/str.
x{i:m}(j) i to m-th part of j-th char/str.
strcat(x1,x2) Concatenate characters/strings.
char(x) Create character from numeric array.
strfind(x1,ptrn) Search x1 char/str for ptrn pattern.
strjoin(x) Construct char array using xchar/str
elements.
lower(x) Convert char/str to lowercase.
upper(x) Convert char/str to uppercase.
strcmp(x1,x2) Compare char/str of x1 and x2.
strcmpi(x1,x2) Compare char/str of x1 and x2 (case
insensitive).
split(x,dl) Split strings in string array (x) at dl
delimiters.
strsplit(x,dl) Split xchar/str at dl delimiters.
sprintf(‘fmt’,x1,x2) Format data based on fmt structure.
strvcat(x1,x2) Vertically concatenate x1 and x2
(ignore spaces).
Regular expression
regexp/regexpi(str,exp) Search exp pattern in str char/
str (case sensitive/insensitive).
regexprep(str,exp,rpc) Replace str which matches exp with
the rpc.
regexptranslate(type,str) Translate str to regular expression
by type as type of translation.
‘IS*’ functions
Return true where:
isnan(X) Elements of Xwhich are NaN.
isnumeric(X) Elements of Xwhich are numeric.
isinf(X) Elements of Xwhich are infinite.
isinteger(X) Elements of Xwhich are integer.
isfloat(X) Elements of Xwhich are floating-point.
isbetween(X,a,b) Elements of Xwhich are between aand b
(date and time).
ismember(X,B) Elements of Xwhich are found in B.
ismissing(X,B) Elements of Xwhich are missing.
Return true if:
isvector(X) X is a vector.
ismatrix(X) X is a logical array.
isstring(X) X is a string (char).
iscell(X) X is a cell array.
iscellstr(X) X is a cell array of strings.
isstruct(X) X is a structure.
istable(X) X is a table.
islogical(X) X is a logical array.
isscalar(X) X is a scalar (size=[1,1]).
isreal(X) There isn’t imaginary value in X.
isrow(X) X is a row vector.
iscolumn(X) X is a column vector.
isdiag(X) X is a lower diagonal matrix.
istril(X) X is a lower triangular matrix.
istriu(X) X is a upper triangular matrix.
isdir(X) X is directory (folder).
MATLAB Cheat Sheet for Data Science - London School of Economics. 2 of 9
‘IS*’ functions...
isequal(X,B) X is equal to B.
isequaln(X,B) X is equal to B(NaN values are equal).
issorted(X) X elements are in ascending order.
isvarname(X) X is a valid MATLAB variable name.
Convert functions
num2str(x) Convert numeric array (x) to char array.
num2cell(x) Convert numeric array (x) to cell array.
num2int(x) Convert numeric array (x) to signed integer.
num2hex(x) Convert numeric array (x) to IEEE
hexadecimal string.
str2num(x) Convert char array (x) to numeric array.
str2mat(x) Convert char/str array (x) to matrix.
str2double(x) Convert char/str array (x) to double
precision.
str2func(x) Convert char array (x) to function handle.
cell2mat(x) Convert cell array (x) to matrix.
cell2table(x) Convert cell array (x) to table.
cell2struct(x) Convert cell array (x) to structure array.
cellstr(x) Convert array xto cell array.
mat2str(x) Convert matrix (x) to char array.
mat2cell(x) Convert matrix (x) to cell array.
table2cell(x) Convert table (x) to cell array.
table2array(x) Convert table (x) to homogeneous array.
table2struct(x) Convert table (x) to structure array.
struct2cell(x) Convert structure array (x) to cell array.
struct2table(x) Convert structure array (x) to table array.
int2str(x) Convert integer (x) to char array.
datenum(x) Convert date and time to a number.
datestr(x) Convert date and time to string.
Programming
Script vs. Function vs. Live script
Script M-files: Contain a list of commands that MATLAB
simply executes in order. They are useful to batch simple
sequences of commonly used commands together.
Function M-files: Can be executed by specifying some
inputs and return some desired outputs.
Live scripts: Contain MATLAB codes, embedded outputs,
formated texts, equations, and images together in a single
environment.
* Add comment: To put a comment within a line, type %
followed by the comment in MATLAB command window,
MATLAB script, or live script enviroment.
% This is a comment line.
x=2; %This is a comment.
y=3;
User-defined functions
Function structure: in1 and in2 are function inputs and
out1 and out2 are function outputs.
function [out1,out2] = fun_name(in1,in2)
...
end
Anonymous function structure: @operator creates a
function handle.
f = @(x)(x.^2+exp(x))
%i.e. f(2) returns 11.3891.
Return: return forces MATLAB to return control to the
invoking function before reaching to the end of that function.
Flow control
If statement: An if statement can be followed by an (or
more) optional elseif and an else statement, which is useful
to test various condition.
if (Condition_1)
MATLAB Commands
elseif (Condition_2)
MATLAB Commands
else
MATLAB Commands
end
Switch statement: Evaluate a statement and selection one
of the cases based on this evaluation.
switch (statement)
case (value1)
MATLAB Commands
case (value2)
MATLAB Commands
end
While loop statement: Repeat the commands while
condition holds.
while (Condition)
MATLAB Commands
end
For loop statement: Loop from ato bin steps of sin the
variable i.
for i = a:s:b
MATLAB Commands
end
Break: break terminates the execution of for or while loop.
Code lines after the break do not execute. In nested loops,
break exits from the loop in which it mentions.
Continue: continue passes control to the next iteration of for
or while loop. In nested loops, continue passes the iteration in
which it mentions.
Errors
Common errors
Error using *: inner matrix dimensions must agree.
The *operator performs matrix multiplication, where an NxM
matrix is multiplied by an MxP matrix, resulting in an NxP
matrix. Use .* instead of *to perform the element-wise
multiplication.
Index exceeds matrix dimensions.
This error arises when you try to reference an element that
doesn’t exist. Some useful functions to check sizes and number
of elements are numel(),size(), and length().
The expression to the left of the equals sign is not a
valid target for an assignment.
This error message arises because of misuse of the =and ==
operators. The =operator does an assignment, and the ==
operator does a logical test for equality.
Subscripted assignment dimension mismatch.
This error message arises because of an attempt to assign a
vector or matrix into a compartment that it does not fit in.
Matrix dimensions must agree.
This error message arises when you try to do operations on
matrices with different dimensions. For example A+B when Ais
2×2 and Bis 2 ×3.
Subscript indices must either be real positive integers
or logicals.
This error message arises because of indexing problem. For
example A(1) is the first element in a matrix not A(0) (like
other programming languages).
MATLAB Cheat Sheet for Data Science - London School of Economics. 3 of 9
Handling errors
Try, catch statement: try a statement, if it returns an
error, catch it and do another statement.
try
statements
catch expression
statements
end
error(‘msg’) Display message and abort function.
warning(‘msg’) Display warning message.
assert(‘msg’) Throw error if condition is false.
st=MException(ID...
,txt) Capture information of a specific
error and save it in the st object.
Parallel computing (CPU & GPU)
CPU:
parpool(size) Create a new parallel pool (size is number of
CPU workers).
gcp Return the current parallel pool.
ticBytes/...
tocBytes(gcp) Start/stop calculation of the bytes transferred
to the workers.
batch(‘scr’) Run a script or function on a worker.
gather(A) Transfer distributed array to local workspace.
parfor: Replace for with parfor to execute code on CPU
workers or cores without any guaranteed order.
parfor i = a:s:b
MATLAB Commands
end
spmd: Execute code in parallel on workers of a pool.
spmd
statements
spmd
parfeval: Directly execute a defined function on a specified
worker.
p=gcp(); % Return current MATLAB pool.
f=parfeval(p,@sum,4,3); % Parallel execution of 4+3.
‘distributed’: Partition listed functions out among the
workers in a pool: zeros(5,5,‘distributed’),ones,false,
true,NaN,inf,eye,rand,randi, and randn.
‘codistributed’: Access the arrays distributed among the
workers in a pool: zeros(5,5,‘codistributed’), etc.
GPU:
gpuDevice(idx) Select GPU device specified by idx.
gpuArray(x) Copy xarray to GPU.
arrayfun(func,A) Apply function to elements of Aon GPU.
bsxfun(func,A,B) Apply an element-wise binary operation
specified by func to Aand Bon GPU.
gather(A) Transfer gpuArray to local workspace.
existsOnGPU(x) Determine if xis stored in GPU.
A basic calculation on GPU:
W=rand(5,‘single’); % Basic random numbers.
GD=gpuArray(W); % Send W to GPU.
GO=GD.*GD; % Execute multiplication on GPU.
Plotting & Figures
figure Open up a new figure window.
axis normal Default axis limits and scaling behavior.
axis tight Force axis to be equaled to data range.
axis equal Force axis to be scaled equally.
axis square Axis lines with equal lengths.
axis fill Lengths of each axis line fill the rectangle.
title(‘Title’) Add a title at the top of the plot.
xlabel(‘lbl’) Label the x axis as lbl.
ylabel(‘lbl’) Label the y axis as lbl.
zlabel(‘lbl’) Label the z axis as lbl.
legend(‘v’,‘w’) Add label to vand wcurves.
grid on/off Include/Omit a grid in the plot.
box on/off Display/hide the box outline around axes.
datetick(‘x’,fm) Date formatted tick labels (fm is format).
xtickformat(fm) X-axis label format.
ytickformat(fm) Y-axis label format.
xlim([min,max]) X-axis limits from min to max.
ylim([min,max]) Y-axis limits from min to max.
zlim([min,max]) Z-axis limits from min to max.
hold on/off Allow/prevent plotting on the same graph.
text(x,y,text) Add text to a point (xand yare scalars
in data units).
fn is a function:
fplot(fn,rn) Plot a 2-D plot using fn over rn range.
fmesh(fn,rn) Plot a 3-D mesh using fn over rn range.
fsurf(fn,rn) Plot a 3-D surface using fn over rn range.
fcontour(fn,rn) Plot contour using a function (fn) over
rn range.
plot(x,y) Plot yversus x(have same length).
plot(y) Plot y, with 1,2,3,... as the xaxis.
plot(x,f(x)) If fis a function, plot the points.
Input arguments:
Line styles: {-/:/-./-- }.
Markers: o: Circle / +: Plus sign / *: Asterisk
/.: Point / x: Cross / s: Square
/d: Diamond / p: Pentagram / h:
Hexagram / ^: Upward triangle.
Colors: y: Yellow / m: Magenta / c: Cyan
/r: Red / g: Green / b: Blue
/w: White / k: Black.
Name-value pair arguments:
Color Line color.
LineStyle Line style.
LineWidth Line width.
Marker Marker symbol.
MarkerIndices Indices of marker data points.
MarkerEdgeColor Marker outline color.
MarkerFaceColor Marker fill color.
MarkerSize Size of marker.
loglog(x,y) Logarithmic x and y axes.
semilogx(x,y) Logarithmic x axis.
semilogy(x,y) Logarithmic y axis.
MATLAB Cheat Sheet for Data Science - London School of Economics. 4 of 9
plot3(x,y,z) Three-dimensional analogue of plot.
surf(x,y,z) 3-D shaded surface plot.
mesh(x,y,z) 3-D mesh surface plot.
histogram(y) Histogram plot.
histfit(y) Histogram plot with distribution fit.
polarhistogram(y) Histogram plot (polar coordi-
nates).
scatter(x,y) 2-D scatter plot by xand y.
scatter(x,y,z) 3-D scatter plot by x,y, and z.
gscatter(x,y,group) 2-D scatter plot of xand y
by group.
bar(y) Bar plot.
bar3(y) 3-D bar plot.
pie(y)/pie3(y) 2-D/3-D pie plot.
MATLAB Cheat Sheet for Data Science - London School of Economics. 5 of 9
area(y) 2-D area plot.
polarplot(theta,rho) Polar plot (theta: angle,
rho: radius).
compass(x) Compass plot (arrows from center).
boxplot(y) Box plot.
candle(y) Candlestick chart.
highlow(h,l,o,c) Plot h(high), l(low), o(open),
and c(close) prices of an asset.
pointfig(y) Point and figure chart.
dendrogram(tree) Dendrogram plot by tree.
contour(y) Contour plot of matrix y.
subplot(a,b,c) For multiple figures in a plot
(a/b: number of rows/columns, c:
selected plot).
Data science
Neural network
nnstart Neural network app (GUI).
patternnet(s,t,p) A pattern recognition (classification) network
with s,t, and pas number of hidden layers,
train, and performance function.
feedforwardnet(s,t) An approximation (regression) network with s
and tas number of hidden layers and train
function.
fitnet(s,t) Function fitting network with sand tas
number of hidden layers and train function.
cascade...
forwardnet(s,t) An approximation (regression) network with s
and tas number of hidden layers and train
function.
selforgmap Design a self-organizing map.
competlayer(nc) Design a competitive layer network with nc as
number of classes.
network Design a custom neural network with different
properties.
view(net) View a designed neural network.
train(net,i,o) Train a network using iand oas input and
output.
MATLAB Cheat Sheet for Data Science - London School of Economics. 6 of 9
predict(net,i) Predictions for ias input by net.
perform(net,i,o) Calculate network performance using i
and oas input and output.
learning functions:
trainlm Levenberg-Marquardt backpropagation.
trainbr Bayesian regularization backpropagation.
trainrp Resilient backpropagation.
trainscg Scaled conjugate gradient backpropagation.
trainbfg BFGS quasi-Newton backpropagation.
traincgb Conjugate gradient backpropagation with
Powell-Beale restarts.
traincgp Conjugate gradient backpropagation with
Polak-Ribire updates.
traincgf Conjugate gradient backpropagation with
Fletcher-Reeves updates.
traingda Gradient descent with adaptive learning rate
backpropagation.
traingdx Gradient descent with momentum and adaptive
learning rate backpropagation.
traingdm Gradient descent with momentum backpropagation.
trainru Unsupervised random order weight/bias training.
trainr Random order incremental training.
trains Sequential order incremental training.
learncon Conscience bias learning function.
learnk Kohonen weight learning function.
learnis Instar weight learning function.
learnos Outstar weight learning function.
Transfer functions:
tansig Hyperbolic tangent sigmoid transfer function.
radbas Radial basis transfer function.
radbasn Normalized radial basis transfer function.
logsig Log-sigmoid transfer function.
tribas Triangular basis transfer function.
purelin Linear transfer function.
satlin Saturating linear transfer function.
poslin Positive linear transfer function.
satlins Symmetric saturating linear transfer function.
hardlim Hard-limit transfer function.
hardlims Symmetric hard-limit transfer function.
elliotsig Elliot symmetric sigmoid transfer function.
elliot2sig Elliot 2 symmetric sigmoid transfer function.
softmax Soft max transfer function.
compet Competitive transfer function.
Performance functions:
mse Mean squared normalized error performance
function.
mae Mean absolute error performance function.
sae Sum absolute error performance function.
sse Sum squared error performance function.
crossentropy Cross entropy performance function.
Input/Output process functions:
mapminmax Normalize inputs/outputs between -1 and 1.
mapstd Zero mean and unity variance normalization.
processpca Principal component analysis for input.
removecons...
tantrows Remove constant inputs/outputs.
fixunknowns Process unknown inputs.
Plots:
ploterrhist Plot error histogram.
plotregression Plot linear regression.
plotfit Plot function fit.
plotperform Plot network performance.
plottrainstate Plot training state values.
plotconfusion Plot classification confusion matrix.
plotroc Plot receiver operating characteristic.
plotsomtop Plot self-organizing map topology.
plotsomhits Plot self-organizing map sample hits.
plotsomnc Plot self-organizing map neighbor
connections.
plotsomnd Plot self-organizing map neighbor distances.
plotsomplanes Plot self-organizing map weight planes.
plotsompos Plot self-organizing map weight positions.
Basic Neural Network implementations (classification &
regression):
%% Classification:
[i,o]=iris_dataset; % Import iris dataset.
nt=patternnet(5); % Design a netwrok.
i_w=nt.IW; % Store initial input weights.
%% Regression:
[i,o]=simplefit_dataset; % Import a sample dataset.
nt=feedforwardnet(10); % Design a network.
nt.performFcn=‘mae’; % Change performance func.
nt.inputs{1}.processFcns={‘processpca’}; % PCA for
input 1.
[nt,tx]=train(nt,i,o); % Train the network.
view(nt) % Show the network.
y=nt(i); % Insert input into the network.
perf=perform(nt,o,y); % Calculate performance.
plotconfusion(c,y); % Plot confusion matrix.
plotperform(tx); % Plot network performance.
Support vector machines/regression
Support vector machines:
classi...
ficationLearner Open classification learner app (GUI).
fitcsvm(i,o) Train SVM with ias input and oas binary
output for low or moderate dimensional data.
fitclinear(i,o) Train linear SVM with ias input and oas binary
output for high dimensional data.
fitcecoc(i,o) Train SVM with ias input and oas multi-class
output by error-correcting output codes model.
fitSVM...
Posterior(svm_m) Return trained SVM that contains the estimated
score transformation function (svm_m is trained
model).
templateSVM
/Linear/ECOC SVM/Linear SVM/Error-correcting output
templates.
predict(svm_c,i) Predict class labels using svm_c trained model
and ias predictor data.
Support vector regression:
regres...
sionLearner Open regression learner app (GUI).
fitrsvm(i,o) Train SVR with ias input and oas output for
low or moderate dimensional data.
fitrlinear(i,o) Train linear SVR with ias input and oas output
for high dimensional data.
predict(svm_r,i) Predict response using svm_r trained model and
ias predictor data.
Basic Support Vector Machines & Support Vector Regression
implementations:
%% Classification:
load fisheriris; % Load iris dataset.
i=meas; % Input data.
o=species; % Output data.
sz=numel(o); % Sample size.
trn_r=1:1:sz-40; % Train range.
tst_r=sz-39:1:sz; % Test range.
svm_c=fitcecoc(i(trn_r,:),o(trn_r,:)); % Train SVM.
svm_cl=crossval(svm_c); % Cross-validation.
svm_c_loss=kfoldLoss(svm_cl); % Error estimation.
pr_out=predict(svm_c,i(tst_r,:)); % Prediction.
%% Regression:
[i,o]=simplefit_dataset; % Import a sample dataset.
svm_r=fitrsvm(i’,o’); % Train SVR.
svm_r_loss_1=resubLoss(svm_r); % Resubstitution loss.
conv=Mdl.ConvergenceInfo.Converged; % Convergance info.
nml=Mdl.NumObservations; % Number of observations.
MATLAB Cheat Sheet for Data Science - London School of Economics. 7 of 9
Deep learning
Autoencoder:
dp1=train...
Autoencoder(i,hs) Train an Autoencoder.
encode(dp1,x) Encode input data (x) using dp1.
decode(dp1,v) Decode encoded data (v) using dp1.
network(dp1) Convert Autoencoder to network object.
plotWeights(dp1) Plot encoder weights of trained Autoencoder.
view(autoenc) View Autoencoder structure.
stack(dp1,dp2) Stack encoders together.
predict(dp1,in) Predict response using dp1 trained model and
in as predictor data.
Convolutional neural network:
op=training...
Options(sv,ops) Training options (ops) of a network (sv: solver
name).
dp2=trainNet...
work(i,o,lyr,op) Train a convolutional network using iand oas
predictor and response variables (op: options).
activations(...
dp2,i,lyr) Network activations for lyr layer using iand
dp2 as data and trained network.
predict(dp2,i) Predict response using dp2 model and ias
model and predictor data.
classify(dp2,i) Classify data using dp2 model and ias
predictor data.
importCaffe...
Network Import pretrained networks models (Caffe).
importCaffe...
Layers Import network layers (Caffe).
dp3=alexnet Pretrained AlexNet network.
dp4=vgg16 Pretrained VGG-16 network.
dp5=vgg19 Pretrained VGG-19 network.
Layers:
imageInputLayer Image input layer.
reluLayer Rectified linear unit layer.
convolution2dLayer Create 2-D convolutional layer.
maxPooling2dLayer Max pooling layer.
fullyConnectedLayer Fully connected layer.
averagePooling
2dLayer Average pooling layer.
crossChannel...
NormalizationLayer Channel-wise local response
normalization layer.
softmaxLayer Softmax layer.
dropoutLayer Dropout layer.
classificationLayer Classification output layer.
regressionLayer Regression output layer.
Basic Autoencoder & Convolutional Neural Network
implementations:
%% Autoencoder:
dt=abalone_dataset; % Load Abalone dataset.
dp1=trainAutoencoder(dt); % Train network.
dt_p=predict(dp1,dt); % Reconstruction.
mse_er=mse(dt-dt_p) % Calculate MSE.
%% Convolutional neural network:
load lettersTrainSet; % Load train data.
load lettersTestSet;; % Load test data.
lyrs=[imageInputLayer([21 21 1]);
convolution2dLayer(8,15);
reluLayer();
fullyConnectedLayer(2);
softmaxLayer();
classificationLayer()];
ops=trainingOptions(’sgdm’); % Settings.
rng(‘default’) % For reproducibility.
nt=trainNetwork(XTrain,TTrain,lyrs,ops)
% Train network.
o_t=classify(nt,XTest); % Classify test data.
r_t=testDigitData.Labels; % Real test labels.
acc=sum(o_t==r_t)/numel(r_t); % Test accuracy.
Decision tree
dt1=fitctree(i,o) Fit a binary classification decision tree to
iand oas predictor and response.
dt2=fitrtree(i,o) Fit a binary regression decision tree to i
and oas predictor and response.
templateTree Design a decision tree template.
dt3=fitc...
ensemble(i,o) Fit ensemble of classification decision
trees to iand oas predictor and
response.
dt4=fitr...
ensemble(i,o) Fit ensemble of regression decision trees
to oand ias predictor and response.
templateEnsemble Design an ensemble model template.
dt5=Tree...
Bagger(nt,i,o) Fit bag of decision trees to iand oas
predictor and response (nt: number of
trees.).
md:dt1,dt2,dt3,dt4, or dt5.
predict(md,in) Predict response using mt fitted model
and in as predictor data.
mdo:dt3,dt4, or dt5.
oobPredict(mdo) Predict out-of-bag response of mdo.
Two basic Decision Trees implementations:
%% A simple classification decision tree.
load fisheriris; % Load iris dataset.
i=meas; % Input data.
o=species; % Output data.
dt1=fitctree(i,o,‘CrossVal’,‘on’); % Fit decision tree
to data using cross validation.
view(dt.Trained{1},‘Mode’,‘graph’); % Graph.
er=kfoldLoss(dt1) % Cross validation error rate.
%% An ensemble learner.
dt2=fitcensemble(i,o,‘Method’,‘Subspace’); Fit an ensemble
using Subspace method.
er=resubLoss(dt2); % Resubstitution loss.
Linear/Nonlinear regression
Linear:
fitlm(x,y) Fit linear regression to xand yas input and
response.
fitglm(x,y) Fit generalized linear regression to xand yas
input and response.
stepwiselm(x,y) Fit stepwise linear regression to xand yas input
and response.
stepwiseglm(x,y) Fit generalized stepwise linear regression to xand
yas input and response.
regress(x,y) Fit multiple linear regression to xand yas input
and response.
fitrlinear(x,y) Fit linear regression to xand yas input and
response for high dimensional data.
robustfit(x,y) Fit robust linear regression to xand yas input
and response for high dimensional data.
mvregress(x,y) Fit multivariate linear regression to xand yas
input and response.
fitlme(tb,fml) Fit linear mixed-effects model for tb and fml as
data table and specified formula.
fitglme(tb,fml) Fit generalized linear mixed-effects model for tb
and fml as data table and specified formula.
ls=lasso(x,y) Fit regularized least-squares regression to xand y
as input and response using lasso or elastic net
algorithms.
ls=lassoglm(x,y) Fit regularized least-squares regression to xand y
as input and response using lasso or elastic net.
algorithm (generalized linear model regression).
lasso(ls) Trace plot of lasso fitted model.
ridge(y,x,k) Fit multilinear ridge regression to x,y, and kas
input, response, and ridge parameters.
plsreg...
ress(x,y,nc) Fit Partial Least-Squares (PLS) regression to x
and yas input and response(nc: PLS components).
MATLAB Cheat Sheet for Data Science - London School of Economics. 8 of 9
Linear: ...
mnrfit(x,y) Fit multinomial logistic regression to xand y
as input and response.
glmfit(x,y) Fit generalized linear model regression to x
and yas input and response.
predict(lr,xn) Predict response using lr trained model and
xn as predictor data.
display(mdl) Display fitted model.
Nonlinear:
fitnlm(x,y,...
mdf,beta) Fit specified model of mdf for x,yand beta
as input, response, and coefficients.
nlinfit(x,y,...
mdf,beta) Fit specified model of mdf for x,yand beta
as input, response, and coefficients.
mnrfit(x,y,...
gr,v,mdf,beta) Fit nonlinear mixed-effects regression for x,
y,gr,v,mdf, and beta as input, response,
groups, predictor (take the same value in a
group), function, and initial estimates for
fixed effects.
nlmefitsa(--) Fit nonlinear mixed-effects model with
stochastic EM algorithm using above inputs.
fitrgp(x,y) Fit a Gaussian Process Regression (GPR)
model to xand yas input and response.
Basic linear & nonlinear regression implementations:
%% Linear regression:
load carsmall; % Load carsmall dataset.
x=[Weight,Acceleration]; % Input data.
y=MPG; % Response data.
lm=fitlm(x,y); % Fit linear regression.
display(lm); % Display reports.
plot(lm); % Scatter plot.
plotAdjustedResponse(lm,1); % Adjusted response plot
of variable 1.
predict(lm,x); % Reconstruct response.
%% Nonlinear regression:
mdf=@(b,x)b(1)+b(2)*x(:,1);
% Model function.
bt=[-50 500]; % Beta values.
nl=fitnlm(x(:,1),y,mdf,bt); % Fit nonlinear
regression.
display(nl); % Display reports.
predict(nl,x(:,1)); % Reconstruct response.
Clustering
linkage(x) Agglomerative hierarchical cluster tree for
xas data.
clusterdata(x,cf) Agglomerative clusters for xand cf as
data and cutoff.
kmeans(x,k) K-means clustering using xand kas data
and number of clusters.
findcluster Fuzzy clustering tool (GUI).
subclust(x,rng) Fuzzy subtractive clustering using xand
rng as data and cluster influence range.
fcm(x,k) Fuzzy c-means clustering using xand k
as data and number of clusters.
kmedoids(x,k) Kmedoids clustering using xand kas
data and number of clusters.
ts=KDTree...
Searcher(x) Grow kd-tree using xas data.
createns(x) Create object for growing kd-tree using x
as data.
ts=Exhaustive...
Searcher(x) Prepare exhaustive nearest neighbors
searcher using xas data.
knnsearch(ts,y) Search for the nearest neighbor in ts to
each point in y.
range...
search(ts,y,r) Find all neighbors within specified
distance in ts to each point in y(r:
radius around each ts to each point).
Two basic clustering models implementations:
% k-means clustering.
load fisheriris; % Load iris dataset.
data=meas(:,1:2); % Select data.
[inx,c]=kmeans(data,3); % k-means clustering.
% inx: cluster indices, c: cluster centers.
% Fuzzy c-means clustering.
[cnt,u]=fcn(data,3); % Fuzzy c-means clustering.
% cnt: cluster centers, u: fuzzy partition matrix.
Dimension reduction/feature selection
c=pca(x) Principal component analysis of x.
c=ppca(x,m) Probabilistic principal component analysis
of xand mas data and number of
components.
biplot(c) Biplot of the PCA coefficients.
pcacov(w) Principal component analysis on was
covariance matrix.
pcares(x,d) Residuals from principal component analysis for x
and das data and number of dimensions.
f=factoran(x,m) Factor analysis of xand mas data and number of
factors.
rotatefactors(f) Rotate factor loadings.
sequentialfs...
(fun,i,o) Sequential feature selection using i,o, and fun as
input, predictor, and function handle that defines
criterion.
relieff(i,o,k) ReliefF algorithm attributes importance extraction
using i,o, and kas input, predictor, and number of
neighbors.
Cross-validation
c=cvpartition...
(o,‘KFold’,k) K-fold cross-validation (o: predictor).
repartition(c) Data repartition for cross-validation.
crossval(fun,x) Loss estimate of cross-validation for the function
fun and xas data.
training(c,ix) Training indices of cross-validation for repetition ix.
test(c,ix) Test indices of cross-validation for repetition ix.
testcholdout...
(y1,y2,yr) Compare predictive accuracies of two classification
models (McNemar test) using y1,y2, and yr as first
model output, second model output, and true labels.
testckfold...
(c1,c2,x1,x2) Compare predictive accuracies of two classification
models by paired F cross-validation test using c1,
c2,x1, and x2 as first model, second model,
first data table and second data table.
Copyright c
2017 / MATLAB R2017a
Ali Habibnia a.habibnia@lse.ac.uk
Eghbal Rahimikia erahimi@ut.ac.ir
Download page: http://personal.lse.ac.uk/habibnia/
* Feel free to send your feedback.
MATLAB Cheat Sheet for Data Science - London School of Economics. 9 of 9
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.