回声状态网络ESN

dataset = csvread('D:\matlab2016a\data.csv');
x = dataset(:,1:3);
y = dataset(:,4);

x_train = x(1:145,:);
y_train = y(1:145);
x_test = x(146:end,:);
y_test = y(146:end);

[samples,inputnum] = size(x_train);
hiddennum = 10;
outputnum = 1;

w_in被随机数生成器生成；
w_state的状态权重矩阵通过随机数生成器计算得出；
w_back的输出权重矩阵同样由随机数生成器产生；

s(:,1) = zeros(hiddennum,1);
for i=1:samples
s(i+1,:)=tanh(w_inx_train(i,:)' + w_states(i,:)' + w_back*y_train(i));
end
s=s(2:end,:);
w_out=(inv(s'*s)s'y_train)';

**for i = 1:samples
predict_train(i) = w_out*s(i,:)';