CSC 367 2.0 Mathematical Computing
Assignment 3 Radial Basis Functions
AS2010377 M.K.H.Gunasekara
Special Part 1 Department of Computer Science UNIVERSITY OF SRI JAYEWARDENEPURA
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
Table of Contents -
Introduction ............................................................................................................................................ 2 Methodology........................................................................................................................................... 3 Implementation ...................................................................................................................................... 5 Results ..................................................................................................................................................... 6 Discussion.............................................................................................................................................. 10 Appendices............................................................................................................................................ 11
1|Page
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
Introduction Neural Networks offer a powerful framework for representing nonlinear mappings from several inputs to one or more outputs. An important application of neural networks is regression. Instead of mapping the inputs into a discrete class label, the neural network maps the input variables into continuous values. A major class of neural networks is the radial basis function (RBF) neural network. We will look at the architecture of RBF neural networks, followed by its applications in both regression and classification. In this report Radial Basis function is discussed for clustering as unsupervised learning algorithm. Radial basis function is simulated to cluster three flowers in a given data set which is available in http://archive.ics.uci.edu/ml/machine-learning-databases/iris/iris.data.
2|Page
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
Methodology Radial Basis Function
Figure 01 : One hidden layer with Radial Basis Activation Functions Radial basis function (RBF) networks typically have three layers 1. Input Layer 2. A hidden layer with a non-linear RBF activation function 3. Output Layer Where N is the number of neurons in the hidden layer, is the center vector for neuron i, and is the weight of neuron i in the linear output neuron. Functions that depend only on the distance from a center vector are radially symmetric about that vector, hence the name radial basis function. In the basic form all inputs are connected to each hidden neuron. The norm is typically taken to be the Euclidean distance and the radial basis function is commonly taken to be Gaussian Function (
)
(
‖
‖
)
------ (1)
There are some other Radial Basis functions Logistic Basis Function ( )
( )
Multi-quadratics ( ) √
3|Page
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
Input nodes connected by weights to a set of RBF neurons fire proportionately to the distance between the input and the neuron in the weight space
The activation of these nodes is used as inputs to the second layer. The second layer (output layer) is treated as a simple Perceptron network Training the RBF Network This can be done positioning the RBF nodes and using the activation of RBF nodes to train the linear outputs. Positioning RBF nodes can be done in two ways; First method is randomly picking some of the data points to act as basis functions. And the second method is trying to position the nodes so that they are representative of typical inputs, like using k-means clustering algorithm. In Activation function there is standard deviation parameter. One option is, giving all nodes the same size, and testing lots of different sizes using a validation set to select one that works. Alternatively we can select the size of RBF nodes so that the whole space is coved by the receptive fields. So the width of the Gaussian should be set according to the maximum distance between the locations of the hidden nodes (d), and the number of hidden nodes (M) ------ (2)
√
We can use this normalized Gaussian function also. (
‖
(
) ∑
(
‖
‖
) ‖
------ (3) )
Outputs of the RBF Network:
(
‖
‖
)
Training the Perceptron Network We can train Pereceptron Network by using supervised learning method. Therefore we train the MLP Network according to targets.
4|Page
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
Implementation Implementation was done using MATLAB 7.10 (2010). Implementation was done according to following methods 1. 2. 3. 4. 5.
Locate RBF nodes into centers Calculate for the Gaussian function Calculate outputs of the RBF layer – Unsupervised Training Make Perceptron Network for second layer –( I used MLP network without a hidden layer) Train MLP Network according to targets and inputs (inputs are the output of RBF network) – Supervised Training 6. Simulate the network
I have implement RBF Network with different strategies to compare the results
Using Randomly selected centers Using K-Means Cluster centers Using Non-normalized Gaussian function Using Normalized Gaussian function Using SVM for second layer
5|Page
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
Results sepal length 5.1 4.9 4.7 4.6 5 5.4 4.6 5 4.4 4.9 5.4 4.8 4.8 4.3 5.8 5.7 5.4 5.1 5.7 5.1 5.4 5.1 4.6 5.1 4.8 5 5 5.2 5.2 4.7 4.8 5.4 5.2 5.5 4.9 5 5.5 4.9
6|Page
sepal width 3.5 3 3.2 3.1 3.6 3.9 3.4 3.4 2.9 3.1 3.7 3.4 3 3 4 4.4 3.9 3.5 3.8 3.8 3.4 3.7 3.6 3.3 3.4 3 3.4 3.5 3.4 3.2 3.1 3.4 4.1 4.2 3.1 3.2 3.5 3.1
petal length 1.4 1.4 1.3 1.5 1.4 1.7 1.4 1.5 1.4 1.5 1.5 1.6 1.4 1.1 1.2 1.5 1.3 1.4 1.7 1.5 1.7 1.5 1 1.7 1.9 1.6 1.6 1.5 1.4 1.6 1.6 1.5 1.5 1.4 1.5 1.2 1.3 1.5
petal width 0.2 0.2 0.2 0.2 0.2 0.4 0.3 0.2 0.2 0.1 0.2 0.2 0.1 0.1 0.2 0.4 0.4 0.3 0.3 0.3 0.2 0.4 0.2 0.5 0.2 0.2 0.4 0.2 0.2 0.2 0.2 0.4 0.1 0.2 0.1 0.2 0.2 0.1
Expected Target Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa
Actual Output Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa
M.K.H.Gunasekara - AS2010377 4.4 5.1 5 4.5 4.4 5 5.1 4.8 5.1 4.6 5.3 5 7 6.4 6.9 5.5 6.5 5.7 6.3 4.9 6.6 5.2 5 5.9 6 6.1 5.6 6.7 5.6 5.8 6.2 5.6 5.9 6.1 6.3 6.1 6.4 6.6 6.8 6.7 6 5.7 5.5
7|Page
3 3.4 3.5 2.3 3.2 3.5 3.8 3 3.8 3.2 3.7 3.3 3.2 3.2 3.1 2.3 2.8 2.8 3.3 2.4 2.9 2.7 2 3 2.2 2.9 2.9 3.1 3 2.7 2.2 2.5 3.2 2.8 2.5 2.8 2.9 3 2.8 3 2.9 2.6 2.4
1.3 1.5 1.3 1.3 1.3 1.6 1.9 1.4 1.6 1.4 1.5 1.4 4.7 4.5 4.9 4 4.6 4.5 4.7 3.3 4.6 3.9 3.5 4.2 4 4.7 3.6 4.4 4.5 4.1 4.5 3.9 4.8 4 4.9 4.7 4.3 4.4 4.8 5 4.5 3.5 3.8
CSC 367 2.0 Mathematical Computing 0.2 0.2 0.3 0.3 0.2 0.6 0.4 0.3 0.2 0.2 0.2 0.2 1.4 1.5 1.5 1.3 1.5 1.3 1.6 1 1.3 1.4 1 1.5 1 1.4 1.3 1.4 1.5 1 1.5 1.1 1.8 1.3 1.5 1.2 1.3 1.4 1.4 1.7 1.5 1 1.1
Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor
Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa Iris-setosa FALSE Iris-versicolor FALSE Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor FALSE Iris-versicolor Iris-versicolor Iris-versicolor FALSE FALSE Iris-versicolor Iris-versicolor Iris-versicolor
M.K.H.Gunasekara - AS2010377 5.5 5.8 6 5.4 6 6.7 6.3 5.6 5.5 5.5 6.1 5.8 5 5.6 5.7 5.7 6.2 5.1 5.7 6.3 5.8 7.1 6.3 6.5 7.6 4.9 7.3 6.7 7.2 6.5 6.4 6.8 5.7 5.8 6.4 6.5 7.7 7.7 6 6.9 5.6 7.7 6.3
8|Page
2.4 2.7 2.7 3 3.4 3.1 2.3 3 2.5 2.6 3 2.6 2.3 2.7 3 2.9 2.9 2.5 2.8 3.3 2.7 3 2.9 3 3 2.5 2.9 2.5 3.6 3.2 2.7 3 2.5 2.8 3.2 3 3.8 2.6 2.2 3.2 2.8 2.8 2.7
3.7 3.9 5.1 4.5 4.5 4.7 4.4 4.1 4 4.4 4.6 4 3.3 4.2 4.2 4.2 4.3 3 4.1 6 5.1 5.9 5.6 5.8 6.6 4.5 6.3 5.8 6.1 5.1 5.3 5.5 5 5.1 5.3 5.5 6.7 6.9 5 5.7 4.9 6.7 4.9
CSC 367 2.0 Mathematical Computing 1 1.2 1.6 1.5 1.6 1.5 1.3 1.3 1.3 1.2 1.4 1.2 1 1.3 1.2 1.3 1.3 1.1 1.3 2.5 1.9 2.1 1.8 2.2 2.1 1.7 1.8 1.8 2.5 2 1.9 2.1 2 2.4 2.3 1.8 2.2 2.3 1.5 2.3 2 2 1.8
Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica
Iris-versicolor Iris-versicolor FALSE Iris-versicolor Iris-versicolor FALSE Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-versicolor Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica FALSE Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica
M.K.H.Gunasekara - AS2010377 6.7 7.2 6.2 6.1 6.4 7.2 7.4 7.9 6.4 6.3 6.1 7.7 6.3 6.4 6 6.9 6.7 6.9 5.8 6.8 6.7 6.7 6.3 6.5 6.2 5.9
3.3 3.2 2.8 3 2.8 3 2.8 3.8 2.8 2.8 2.6 3 3.4 3.1 3 3.1 3.1 3.1 2.7 3.2 3.3 3 2.5 3 3.4 3
CSC 367 2.0 Mathematical Computing
5.7 6 4.8 4.9 5.6 5.8 6.1 6.4 5.6 5.1 5.6 6.1 5.6 5.5 4.8 5.4 5.6 5.1 5.1 5.9 5.7 5.2 5 5.2 5.4 5.1
2.1 1.8 1.8 1.8 2.1 1.6 1.9 2 2.2 1.5 1.4 2.3 2.4 1.8 1.8 2.1 2.4 2.3 1.9 2.3 2.5 2.3 1.9 2 2.3 1.8
Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica
Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica FALSE Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica Iris-virginica
I found best results using RBF Network with Non-Normalized Gaussian activation function with 9 mismatches. And I found best results using MLP Network with 4 mismatches. MLP Network as Second Layer
Non-Normalized Gaussian function Normalized Gaussian function
Random Center 9
K Means Center 9
11
11
Support Vector Machine as Second Layer
Non-Normalized Gaussian function Normalized Gaussian function
9|Page
Random Center 14
K Means Center 10
14
17
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
Discussion 1. There are some drawbacks of unsupervised center selection in radial basis functions 2. We can use an SVM for the second layer instead of a perceptron but it is not efficient for more than 2 classes classification
10 | P a g e
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
Appendices MATLAB Sourcecode for RBF Network with MLP Network clc clear all % M.K.H. Gunasekara % AS2010377 % Machine Learning % Radial Basis Function [arr tx] = xlsread('data.xls'); Centers=zeros(3,4);
% I found centers as mean of the same cluster values for i=1:50 Centers(1,1)=arr(i,1)+Centers(1,1); Centers(1,2)=arr(i,2)+Centers(1,2); Centers(1,3)=arr(i,3)+Centers(1,3); Centers(1,4)=arr(i,4)+Centers(1,4); end for i=51:100 Centers(2,1)=arr(i,1)+Centers(2,1); Centers(2,2)=arr(i,2)+Centers(2,2); Centers(2,3)=arr(i,3)+Centers(2,3); Centers(2,4)=arr(i,4)+Centers(2,4); end for i=101:150 Centers(3,1)=arr(i,1)+Centers(3,1); Centers(3,2)=arr(i,2)+Centers(3,2); Centers(3,3)=arr(i,3)+Centers(3,3); Centers(3,4)=arr(i,4)+Centers(3,4); end for j= 1:3 Centers(j,1)=Centers(j,1)/50; Centers(j,2)=Centers(j,2)/50; Centers(j,3)=Centers(j,3)/50; Centers(j,4)=Centers(j,4)/50; end Centers
% OR we can use k means algorithms calculate cluster centers k=3; %number of clusters [IDX,C]=kmeans(arr,k); C %RBF centres %Uncomment following line to use k means %Centers=C;
11 | P a g e
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
% distance between hidden nodes %distance between hidden node 1 & 2 dist1= sqrt((Centers(1,1)-Centers(2,1))^2 + (Centers(1,2)-Centers(2,2))^2 + (Centers(1,3)-Centers(2,3))^2 + (Centers(1,4)-Centers(2,4))^2); %distance between hidden node 1 & 3 dist2= sqrt((Centers(1,1)-Centers(3,1))^2 + (Centers(1,2)-Centers(3,2))^2 + (Centers(1,3)-Centers(3,3))^2 + (Centers(1,4)-Centers(3,4))^2); %distance between hidden node 3 & 2 dist3= sqrt((Centers(3,1)-Centers(2,1))^2 + (Centers(3,2)-Centers(2,2))^2 + (Centers(3,3)-Centers(2,3))^2 + (Centers(3,4)-Centers(2,4))^2);
% finding maximum distance maxdist=0; if ( dist1>dist2) & (dist1>dist3) maxdist=dist1; end if ( dist2>dist1) & (dist2>dist3) maxdist=dist2; end if ( dist3>dist1) & (dist3>dist2) maxdist=dist3; end % calculating width sigma= maxdist/sqrt(2*3);
maxdist; % Gaussian %calculating outputs of RBF networks RBFoutput=zeros(150,3); d1=zeros(1,4); Centers; d=zeros(1,3); %Unnormalized method % calculate output for gaussian function %Uncomment following lines (98-106) to use Non-Normalized Activation %functions % for i=1:150 for j=1:3 d(1,j)= (arr(i,1)- Centers(j,1))^2 + (arr(i,2)- Centers(j,2))^2 + (arr(i,3)- Centers(j,3))^2 + (arr(i,4)- Centers(j,4))^2; RBFoutput(i,j)= exp(-(d(1,j)/(2*(sigma^2)))); end end
12 | P a g e
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
% %
%Normalized method %Summation %Uncomment following lines (114-130) to use Gaussian Normalized Activation functions % RBFNormSum=zeros(150,1); % for i=1:150 % for j=1:3 % d(1,j)= (arr(i,1)- Centers(j,1))^2 + (arr(i,2)- Centers(j,2))^2 + (arr(i,3)- Centers(j,3))^2 + (arr(i,4)- Centers(j,4))^2; % RBFNormSum(i,1)= exp(-(d(1,j)/(2*(sigma^2))))+ RBFNormSum(i,1); % end % % d=[0 0 0]; % end % % % calculate output for gaussian function % for i=1:150 % for j=1:3 % d(1,j)= (arr(i,1)- Centers(j,1))^2 + (arr(i,2)- Centers(j,2))^2 + (arr(i,3)- Centers(j,3))^2 + (arr(i,4)- Centers(j,4))^2; % % RBFoutput(i,j)= exp(-(d(1,j)/(2*(sigma^2))))/RBFNormSum(i,1); % end % % d=[0 0 0]; % end
RBFoutput RBFo=RBFoutput.' % making MLP network % T=zeros(1,150); T=[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3] S=[3 1] ; R=[0 1;0 1;0 1]
% used feedforward neural network as MLP [3 1] MLPnet=newff(RBFo,S); MLPnet.trainParam.epochs = 500; MLPnet.trainParam.lr = 0.1; MLPnet.trainParam.mc = 0.9; MLPnet.trainParam.show = 40;
13 | P a g e
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
MLPnet.trainParam.perf = 'mse'; MLPnet.trainParam.goal = 0.001; MLPnet.trainParam.min_grad = 0.00001; MLPnet.trainParam.max_fail=4;
MLPnet = train(MLPnet,RBFo,T); %simulating neural network y=sim(MLPnet,RBFo); output=round(y.'); Target=T.'; compare= [T.' output] count=0; for i=1:150 if(output(i)~=Target(i)) count=count+1; end end Unmatched=count
MATLAB Source code for RBF Network with SVM clc clear all % M.K.H. Gunasekara % AS2010377 % Machine Learning % Radial Basis Function with Support Vector Machine [arr tx] = xlsread('data.xls'); Centers=zeros(3,4);
% I found centers as mean of the same cluster values for i=1:50 Centers(1,1)=arr(i,1)+Centers(1,1); Centers(1,2)=arr(i,2)+Centers(1,2); Centers(1,3)=arr(i,3)+Centers(1,3); Centers(1,4)=arr(i,4)+Centers(1,4); end for i=51:100 Centers(2,1)=arr(i,1)+Centers(2,1); Centers(2,2)=arr(i,2)+Centers(2,2); Centers(2,3)=arr(i,3)+Centers(2,3); Centers(2,4)=arr(i,4)+Centers(2,4); end for i=101:150 Centers(3,1)=arr(i,1)+Centers(3,1); Centers(3,2)=arr(i,2)+Centers(3,2); Centers(3,3)=arr(i,3)+Centers(3,3);
14 | P a g e
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
Centers(3,4)=arr(i,4)+Centers(3,4); end for j= 1:3 Centers(j,1)=Centers(j,1)/50; Centers(j,2)=Centers(j,2)/50; Centers(j,3)=Centers(j,3)/50; Centers(j,4)=Centers(j,4)/50; end Centers
% OR we can use k means algorithms calculate cluster centers k=3; %number of clusters [IDX,C]=kmeans(arr,k); C %RBF centres %Uncomment following line to use k means Centers=C;
% distance between hidden nodes %distance between hidden node 1 & 2 dist1= sqrt((Centers(1,1)-Centers(2,1))^2 + (Centers(1,2)-Centers(2,2))^2 + (Centers(1,3)-Centers(2,3))^2 + (Centers(1,4)-Centers(2,4))^2); %distance between hidden node 1 & 3 dist2= sqrt((Centers(1,1)-Centers(3,1))^2 + (Centers(1,2)-Centers(3,2))^2 + (Centers(1,3)-Centers(3,3))^2 + (Centers(1,4)-Centers(3,4))^2); %distance between hidden node 3 & 2 dist3= sqrt((Centers(3,1)-Centers(2,1))^2 + (Centers(3,2)-Centers(2,2))^2 + (Centers(3,3)-Centers(2,3))^2 + (Centers(3,4)-Centers(2,4))^2);
% finding maximum distance maxdist=0; if ( dist1>dist2) & (dist1>dist3) maxdist=dist1; end if ( dist2>dist1) & (dist2>dist3) maxdist=dist2; end if ( dist3>dist1) & (dist3>dist2) maxdist=dist3; end % calculating width sigma= maxdist/sqrt(2*3);
15 | P a g e
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
maxdist; % Gaussian %calculating outputs of RBF networks RBFoutput=zeros(150,3); d1=zeros(1,4); Centers; %Unnormalized method % calculate output for gaussian function %Uncomment following lines (98-106) to use Non-Normalized Activation %functions d=zeros(1,3); for i=1:150 for j=1:3 d(1,j)= (arr(i,1)- Centers(j,1))^2 + (arr(i,2)- Centers(j,2))^2 + (arr(i,3)- Centers(j,3))^2 + (arr(i,4)- Centers(j,4))^2; RBFoutput(i,j)= exp(-(d(1,j)/(2*(sigma^2)))); end % d=[0 0 0]; end %
%Normalized method %Summation %Uncomment following lines (114-130) to use Gaussian Normalized Activation functions % RBFNormSum=zeros(150,1); % for i=1:150 % for j=1:3 % d(1,j)= (arr(i,1)- Centers(j,1))^2 + (arr(i,2)- Centers(j,2))^2 + (arr(i,3)- Centers(j,3))^2 + (arr(i,4)- Centers(j,4))^2; % RBFNormSum(i,1)= exp(-(d(1,j)/(2*(sigma^2))))+ RBFNormSum(i,1); % end % % d=[0 0 0]; % end % % % calculate output for gaussian function % for i=1:150 % for j=1:3 % d(1,j)= (arr(i,1)- Centers(j,1))^2 + (arr(i,2)- Centers(j,2))^2 + (arr(i,3)- Centers(j,3))^2 + (arr(i,4)- Centers(j,4))^2; % % RBFoutput(i,j)= exp(-(d(1,j)/(2*(sigma^2))))/RBFNormSum(i,1); % end % % d=[0 0 0]; % end
RBFoutput RBFo=RBFoutput.' % making SVM network
16 | P a g e
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
group=cell(3,1) group{1,1}=zeros(150,1); for n=1:150; tclass(n,1)=tx(n,5); end group{1,1}=ismember(tclass,'Iris-setosa') group{2,1}=ismember(tclass,'Iris-versicolor') group{3,1}=ismember(tclass,'Iris-virginica')
[train, test] = crossvalind('holdOut',group{1,1}); cp = classperf(group{1,1}); for i=1:3 %svmStruct(i) = svmtrain(RBFoutput(train,:),group{i,1}(train),'showplot',true); svmStruct(i) = svmtrain(RBFoutput,group{i,1},'showplot',true); end for j=1:size(RBFoutput) for k=1:3 if(svmclassify(svmStruct(k),RBFoutput(j,:))) break; end end result(j) = k; end T=[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3] compare=[T.' result.'] Target=T.' output=result.' count=0; for i=1:150 if(output(i)~=Target(i)) count=count+1; end end Unmatched=count
MATLAB Source Code MLP Network clc clear all % M.K.H. Gunasekara % AS2010377 % Machine Learning % MLP Network [arr tx] = xlsread('data.xls');
17 | P a g e
M.K.H.Gunasekara - AS2010377
CSC 367 2.0 Mathematical Computing
inputs=arr.'; T=[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3] %Multilayer network with hidden layer with 3 nodes MLPnet=newff(inputs,[4 3 1]); MLPnet.trainParam.epochs = 500; MLPnet.trainParam.lr = 0.1; MLPnet.trainParam.mc = 0.9; MLPnet.trainParam.show = 40; MLPnet.trainParam.perf = 'mse'; MLPnet.trainParam.goal = 0.001; MLPnet.trainParam.min_grad = 0.00001; MLPnet.trainParam.max_fail=4;
MLPnet = train(MLPnet,inputs,T); %simulating neural network y=sim(MLPnet,inputs); output=round(y.'); Target=T.'; compare= [T.' output] count=0; for i=1:150 if(output(i)~=Target(i)) count=count+1; end end Unmatched=count
18 | P a g e
