is my neural network properly trained ?

1 view (last 30 days)
Newman
Newman on 17 Jul 2016
Edited: Newman on 19 Jul 2016
Hello i have a feature vector of 12455x400 matrix and out put target of 40x400 matrix .I am training using patternent and fucntion is trainscg for face recognition using local binary pattern. I hve 400 images which have 40 individuals in them. This is my neural network output .
8.333 mis classifications means it had identified almost all other ~92% of images in the test set of 60 images?

Accepted Answer

Greg Heath
Greg Heath on 17 Jul 2016
Yes.
However, isn't there any way to use feature extraction to reduce your input dimension below 12455 ???
Can patternnet really handle vectors that long?
Thank you for formally accepting my answer
Greg
  3 Comments
Greg Heath
Greg Heath on 19 Jul 2016
Inputs that can be approximated by combinations of other inputs should be removed.
In particular, it is those that can be approximated by LINEAR combinations of other inputs that cause the most trouble.
Typically, the larger the number of known input/output PAIRS the better the classification.
Hope this helps.
Thank you for formally accepting my answer
Greg
Newman
Newman on 19 Jul 2016
Edited: Newman on 19 Jul 2016
Sir how to find out those kind of inputs ? Do u know of any filtering techniques for the feature vector ? Or is it done manually.

Sign in to comment.

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!