[SOLVED] matlab Clarify on Assignment-1

Clarify on Assignment-1

Matlab Functions and Tools

Some functions are given. There is no need to
implement them. Please refer to the links in
the next slide.

You can also use other resources, with proper
citation in your report.

Useful Matlab Functions
Nave Bayes classifier

PredictClass = classify(Xtest,Xtrain,Ytrain,diaglinear);
Randomly split data

p = randperm(n,k)
Indices = crossvalind(Kfold, N, K)

plotImages
plotImages(digitsImages, xy_coord, scale, skip);

LLE:
http://www.cs.nyu.edu/~roweis/lle/code.html

ISOMAP:
http://web.mit.edu/cocosci/isomap/isomap.html

LDA-dimension reduction
http://lvdmaaten.github.io/drtoolbox/

decide the
model to learn

n: total # of samples
k: select k samples by permutation

http://www.cs.nyu.edu/~roweis/lle/code.html
http://isomap.stanford.edu/
http://isomap.stanford.edu/
http://homepage.tudelft.nl/19j49/Matlab_Toolbox_for_Dimensionality_Reduction.html
http://homepage.tudelft.nl/19j49/Matlab_Toolbox_for_Dimensionality_Reduction.html

Datasets

Dataset A (record activity sensors):
Sample-feature matrix: fea (19,000 x 81)

Features: readings of 81 sensors
The data is in time-series, given in time order

Missing values
NaN

Outliers
Negative readings are not outliers

Datasets

Dataset B (image data of handwritten digits)
Sample-feature matrix: fea (2066 x 784)

Features: 28 x 28 gray-scale images, in column-wise
Ground truth labels: gnd (2066 x 1)

Labels: 0, 1, 2, 3, 4

Datasets

Dataset C (clinic data)
Sample-feature matrix:

fea (2100 x 21)
Need to be normalized (min-max) before further

processing

Ground truth labels
gnd (2100 x 1)
3 classes: normal(1), suspect(2), pathologic(3)

Q1: Data Cleaning and Preprocessing

Missing values; Outliers
Detect and fix them

Normalization:
Min-max
Z-score

Plot histograms
Observations and comments!

Q2&3: Feature Extraction

Linear methods:
PCA
LDA

Nonlinear methods:
LLE
ISOMAP

Supervised vs. non-supervised dimensionality
reduction

Q4: Feature Selection

Search strategy
SFS
SBS

Objective function
Filter based
Wrapper based

basics

Hello World!
a = 3;
b = 4;
c = a + b

end each statement with semicolon, if you do
not like to see the result in the command
window

basics
arithmetic operators:

addition: A+B
subtraction: A-B
multiplication: A*B
right division: A/B = A*inv(B)
left division: AB = inv(A)*B
power: A^b
transpose: A
colon operator:

to create vectors: a:b
array subscripting: A(:,b)

basics

dot operators (a.k.a element-wise operators)
A.*B, A./B, A.B, and A.^B

relational operators
a b, a>=b, a==b, and a~=b

logical operations
a||b (or),a&&b (and), ~ a(not)

element-wise logical operators
A|B, A&B, ~A

basics

operator precedence
Parentheses
transpose and power
unary plus, unary minus, and logical negation
multiplication(s) and division(s)
addition and subtract

basics

flow control
conditional control

if, else, and elseif
switch and case

loop control
for
while
break
continue

basics

if
if expression1

statements1
elseif expression2

statements2

else
statements3

end

basics

for
for index = values

program statements
end

while
while expression

program statements
end

basics

function definition
function [output_variables] =
fcn_name(input_variables)

the name of a function should be consistent with
the file name

an example

Given the corresponding coefficients of two lines
(ax+by+c=0), calculate the intersection point and
plot the lines on a figure.

Function:
[intersection,Runtime] = myPlot(line1,line2)
Script to call this function:

clear all;clc
load coeffs
line1 = coeff(1,:);
line2 = coeff(2,:);
[intersection,RunTime] = myPlot(line1,line2)

how to access?

have a license
Nexus computers

on campus
remotely

Octave

help

where to look for answers?
Matlab Help
Mathworks website
Online forums
TAs

refs

www.mathworks.com
www.gnu.org/software/octave/
saw.uwaterloo.ca/matlab/