BAM-neural-net

Background

Based on bidirectional heteroassociative memory model by Chartier & Boukadoum, 2006. Learns a mapping function for binary-to-binary vector transformations.

Free parameters

delta

epsilon = f(hetta)

number of trials

number of iterations

Output

y squared error

x squared error

W => Y = WX (threshold > 0 first)

V => X = VY (threshold > 0 first)

Example

X = abs(double(normrnd(0,1,35,5))>0.1);

Y = abs(double(normrnd(10,10,2000,5))>3);

[y_sq_error x_sq_error W V] = linear_associator(X,Y,0.1,64,0.5,5);