CSS485_Perceptron/question4.mat at main · BenYang2002/CSS485_Perceptron · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
% input for 0
input0 = [-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]';
% input for 1
input1 = [-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]';
% input for 2
input2 = [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]';
%output matrix
output_matrix = [1,0,0;0,1,0;0,0,1];
input_matrix = [input0,input1,input2];
% create the perceptron
init_weight_zeros = zeros(3,30);
init_weight_bias = zeros(3,1);
%number_identifier = PerceptronLayer(init_weight_zeros,init_weight_bias,"hardlim");
number_identifier = PerceptronLayer(30,3,"hardlim");
number_identifier = number_identifier.input_setter(input_matrix);
number_identifier = number_identifier.output_setter(output_matrix);
number_identifier = number_identifier.learn();
disp("weight matrix");
neuron1 = reshape(number_identifier.weight_matrix(1,:),5,6);
neuron2 = reshape(number_identifier.weight_matrix(2,:),5,6);
neuron3 = reshape(number_identifier.weight_matrix(3,:),5,6);
disp("weight for neuron1");
disp(neuron1);
disp("weight for neuron2");
disp(neuron2);
disp("weight for neuron3");
disp(neuron3);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%part2%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
accuracy1 = [];
accuracy2 = [];
accuracy3 = [];
for j = 1 : 4
    % this vector keep track of the correct times for three kinds of input
    correct_times = [0,0,0];
    flip_input = zeros(30,3);
    prediction = zeros(3,3);
    test_size = 1000;
    for i = 1 : test_size
        for k = 1:3
            filp_input(:,k) = addNoise(input_matrix(:,k),2 * j);
            prediction(:,k) = number_identifier.forward(filp_input(:,k));
            error = number_identifier.errorLoss(prediction(:,k),output_matrix(:,k));
            if all(error == 0)
                correct_times(k) = correct_times(k) + 1;
            end
        end
    end
    disp("correct rate for flip " + 2 * j + " pixels for pattern 0 is: " + correct_times(1)/test_size);
    disp("correct rate for flip " + 2 * j + " pixels for pattern 1 is: " + correct_times(2)/test_size);
    disp("correct rate for flip " + 2 * j + " pixels for pattern 2 is: " + correct_times(3)/test_size);
    disp(" ");
    accuracy1 = [accuracy1,correct_times(1)/test_size];
    accuracy2 = [accuracy2,correct_times(2)/test_size];
    accuracy3 = [accuracy3,correct_times(3)/test_size];
end
flip_num = [2,4,6,8];
plot(flip_num, accuracy1, "o-","DisplayName","0 pattern");
hold on;
plot(flip_num, accuracy2, "o-","DisplayName","1 pattern");
hold on;
plot(flip_num, accuracy3, "o-","DisplayName","2 pattern");
legend("show");
hold off;