How to crete Input and Target data for Neural Network Training?

I have character dataset for handwriting recognition from 110 users. Each user writes 110 different characters twice. So basically there are 110*110*2 images in total. I wrote a feature extraction code and now have 80 features per image (character). The FE code has created 110 text files representing each character i.e. one file represents one of the 110 characters from all the users. Each file has name of the user and image number followed by 80 features separated by comma. Since there are 110 characters to be recognized, I know there should be 110 classes but I am not sure how to create matrices for input and target so the dimensions match. Could someone please help me, how can solve this problem.

Answer 1

[Greg Heath]

N = 110*110*2 % 24200

[ I N ] = size(input) % [ 80 N ]

[ O N ] = size(target) % [ 110 N ] e.g, [ repmat(eye(110), 1, 220)]

Hope this helps.

Thank you for formally accepting my answer

Greg

[Harsimrat Parmar]

Thank you so much for your reply. It helped me create my input matrix but I am still struggling with the target matrix.

>> whos

Name          Size               Bytes  Class     Attributes
data      22289x84            14978208  double

This is my input matrix. How can I create my target matrix for these dimensions.

Also I am confused about how will the learning take place. A particular row in input matrix must be mapped to a target row (I think) for the NN to learn which class is represented by that row. Please correct me if I am wrong. If this is correct, how will the mapping between the input and target rows be done?

[Greg Heath]

 help vec2ind
 doc vec2ind

 help ind2vec2
 doc ind2vec2

For examples search the NEWSGROUP and ANSWERS using

 patternnet ind2vec vec2ind