how to create input and target for this feature vector to train a neural network

Hi all, I need to classify the malaria infected microscopic images into its respective stages of infection. I have extracted 20 features from 4 images belonging to different stages and formed 4 feature vectors of size 20x1 each. Also i have saved it in workspace as 4 individual feature vectors of size 20x1 each. I don't know how to create input and target for this feature vector so that i can train the neural network. Kindly help

Answer 1

[Greg Heath]

If you have N I-dimensional input column vectors that need to be classified into 1 of c independent classes, the N target vectors are {0,1} c-dimensional unit column vectors from the unit matrix eye(c).

with I = 20 you need N at least, several times larger than I. To be more precice:

For an I-H-c classifier node structure

 Number of training equations 

Ntrneq = 0.7*N*c  % 0.7 is a default value

 Number of unknown weights 

Nw = (I+1)H + (H+1)*c = c +(I+c+1)*10 % H = 10 is a default value

For Nw not to exceed Ntrneq,

           0.7*N*c  >= c +(I+c+1)*10

             2.8*N >= 4 +(20+4+1)*10
                 N >= 91

Obviously you need some combination of increasing N and/or decreasing I*H

Hope this helps.

Thank you for formally accepting my aanswer

Greg