How to create a target matrix for neural network?

Hello!

I extracted features from a database of 894 images. Each image had 500 extracted features and this process resulted in an array of 500x894 (This the training data). The database has 5 classes: class 1 has 128 images; class 2 has 312 images; class 3 has 125 images; class 4 has 180 images; and class 5 has 149 images.

So, how do I create target matrix for the network?

[KSSV]

YOu can give some number to each class...may be class1 as 1, class5 as 5.

[Greg Heath]

ABSOLUTELY NOT !!!

Answer 1

[Greg Heath]

The classic classifier target matrix for c classes contain columns of the unit matrix eye(c).

The corresponding output layer function is SoftMax with unit sum columns representing class posterior probabilities.

See any reputable neural network text

Thank you for formally accepting my answer

Greg