Artificial Neural Network layers -

i have decided try , make reccognition system. , want start pictures of say, 16x16 pixels. 256 input neurons. now, output neurons essensially how many results want, want distinguish letters a, b , c. need 3 output neurons right?

my question is, how can know how many neurons need in hidden layer? , purpose of them again? how many character classes want? say, o , q quite simular, thay both lead 1 hidden layer neuron later tell them appart?

you're right input , output layers.

how can know how many neurons need in hidden layer?

there's no concrete rule says how many units need in hidden layers of neural network. there general guidelines though, i'll quote one of answers on cross validated.

number of input units: dimension of features x(i)

number of output units: number of classes

reasonable default 1 hidden layer, or if > 1 hidden layer, have same number of hidden units in every layer (usually more better, anywhere 1x 4x number of input units).

you asked:

and purpose of them again?

the hidden layer units transform inputs values (using coefficients selected during training) can used output layer.

is how many character classes want? say, o , q quite similar, thy both lead 1 hidden layer neuron later tell them apart?

no, that's not right. number of output units same number of classes want. each output unit correspond 1 letter, , whether or not input image letter (with probability). output unit highest probability 1 select right letter.