I've tried to make an NN for a connect 4 game according to some article i've
found on the net.
My topology is : 126(3 inputs for each position(7x6 board)) input neurons,
126 hidden neurons, 7 output neurons
each input neuron is connected with one hidden neuron and each hidden neuron
is connected with all of the output neurons.
The output neurons produce a value between 0 and 1 and the neuron with the
highest output is used to play the corresponding column.
When I train the network and try to run some values through it(even values
that i trained) i always get the same output no matter what input i give. So
the same column is played over and over.
This doesn't happen when i train just one inputvalue but as soon as i train
more than one it's the same again.
Would this be an error in my train algorithm or is there a problem with my
here is my training function
Thanks in advance.