4-class perceptron classification

Question

Bill Symolon 2018-2-8

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/381416-4-class-perceptron-classification

评论： Bill Symolon 2024-5-20

Good evening, I hope everyone is well. I have a single-layer perceptron that is meant to take in two inputs and provide an output as one of four classifications. I then need to plot the inputs and the hyperplanes dividing the four classes. I'm getting errors when I try to run the code that I suspect are related to the number of outputs I'm trying to get. Here's the code I'm working with:

if true
   % Initialize the Input Space (extended input matrix)
   x = [1 1 1 2 2 -1 -2 -1 -2; 1 1 2 -1 0 2 1 -1 -2];
     % Initialize the extended target vector
     t = [0 1 1 2 2 3 3 4 4];
     % Plot the input locations
     figure
     hold on
     for i=1:length(x)
        if (t(i)==1)
            scatter(x(1,i), x(2,i),'k ', 'filled');
        elseif (t(i)==2)
            scatter(x(1,i), x(2,i),'r ', 'filled');
        elseif (t(i)==3)
            scatter(x(1,i), x(2,i),'b ', 'filled');
        elseif (t(i)==4)
            scatter(x(1,i), x(2,i),'g ', 'filled');
        end
     end
     grid on
     line([0 0], ylim, 'linewidth', 1);  %y-axis
     line(xlim, [0 0], 'linewidth', 1);  %x-axis
     legend('class_1', 'class_2', 'class_3', 'class_4')
     net = perceptron;
     net.trainparam.epochs = 100;    % Set # of training epochs
     net.trainparam.goal = 1e-2;     % Set desired max error
     net.trainparam.lr = 0.1;        % Set desired learning rate
     train(net, x, t);               % Train the perceptron
     predictions = net(x);           % Get data predictions
     net.IW{:};      % Learned weights
     net.b{:};       % Learned biases
     % Equation of a line: w1*x1 + w2*x2 + b = 0
     % Plot the lines 
     plot([-net.b{1}/net.IW{:}(1,1),0],[0,-net.b{1}/net.IW{:}(1,2)])
     plot([-net.b{2}/net.IW{:}(2,1),0],[0,-net.b{2}/net.IW{:}(2,2)])
     hold off;
  end

The trainparam epochs, goal and lr are initial conditions an can change but I don't think that's where the problem is.

I'd appreciate any help you're able to provide. Best regards.

4 个评论
显示 2更早的评论隐藏 2更早的评论

Akash 2024-5-19

can you send the modified code

Bill Symolon 2024-5-20

Dang, asking me to go back 6 years .... it took me a minute to find this :)

% Initialize Variables for the Input Spaces (x)

p1 = [1; 1];

p2 = [1; 2];

p3 = [-2; -1];

p4 = [2; 0];

p5 = [-1; 2];

p6 = [-2; 1];

p7 = [-1; -1];

p8 = [-2; -2];

% Concatenate the input vectors into a single input matrix

x = [p1 p2 p3 p4 p5 p6 p7 p8];

% Initialize target vector

t1 = [0 0 0 0 1 1 1 1];

t2 = [0 0 1 1 0 0 1 1];

% Plot the input locations

figure

hold on;

for i=1:length(x)

if t1(i) == 0 && t2(i) == 0

scatter(x(1,i), x(2,i),'k ', 'filled');

elseif t1(i) == 0 && t2(i) == 1

scatter(x(1,i), x(2,i),'r ', 'filled');

elseif t1(i) == 1 && t2(i) == 0

scatter(x(1,i), x(2,i),'b ', 'filled');

else

scatter(x(1,i), x(2,i),'g ', 'filled');

end

grid on

line([0 0], ylim, 'linewidth', 1); %y-axis

line(xlim, [0 0], 'linewidth', 1); %x-axis

% Create two perceptrons

net1 = perceptron;

net2 = perceptron;

% Set training parameters

net1.trainparam.epochs = 1000; % Set # of training epochs

net2.trainparam.epochs = 1000;

net1.trainparam.goal = 1e-5; % Set desired max error

net2.trainparam.goal = 1e-5;

net1.trainparam.lr = 0.01; % Set desired learning rate

net2.trainparam.lr = 0.01;

% Train the perceptrons

net1 = train(net1, x, t1);

net2 = train(net2, x, t2);

% Get data predictions

pred1 = net1(x);

pred2 = net2(x);

w_net1 = net1.IW{:}; % net1 learned weights

b_net1 = net1.b{:}; % net1 learned biases

w_net2 = net2.IW{:}; % net2 learned weights

b_net2 = net2.b{:}; % net2 learned biases

% Plot the lines

plot([-b_net1/w_net1(1),0],[0,-b_net1/w_net1(2)], 'k-');

plot([-b_net2/w_net2(1),0],[0,-b_net2/w_net2(2)], 'k--');

hold off;

请先登录，再进行评论。

请先登录，再回答此问题。

4-class perceptron classification

4 个评论
显示 2更早的评论隐藏 2更早的评论

回答（0 个）

另请参阅

类别

标签

Community Treasure Hunt

4-class perceptron classification

4 个评论 显示 2更早的评论隐藏 2更早的评论

回答（0 个）

另请参阅

类别

标签

Community Treasure Hunt

4 个评论
显示 2更早的评论隐藏 2更早的评论