Is it necessary to have equal number of samples(observations) belonging to the two classes for training an NN classifier for a two-class problem?

Hi!

This is a rather basic question, but I want to know if it is necessary to have exactly equal number of samples(observations) belonging to the two classes for training an NN classifier for a two-class problem. If there is a slight imbalance, say by up to 10%, how would it affect the classifier performance?

Thanks.

Answer 1

[Greg Heath]

Hmm, the lengthy dissertation I wrote yesterday seems to have disappeared. I'm too lazy to go through that again. So, the short version:

1. 10% should work

2. Search NEWSGROUP and ANSWERS for

 greg unbalanced

Hope this helps.

Thank you for formally accepting my answer

Greg

[TS Sharma]

Thank you Greg!