Confusion matrix neural network plot interpretation

I would greatly appreciate if someone could help me interpret the confusion matrix below. I don't understand the red, green and grey boxes.Also, what is the difference between test, validation, training and all confusion matrix. Thanks in advance!

Answer 1

[Greg Heath]

By default, the data is divided into three subsets (trn/val/tst):

total = training + validation + testing

You can think of it two different ways:

1. total      = designing + testing
   designing  = training + validation
2. total      = training + nontraining
  nontraining = validation + testing

TRAINING SUBSET: Design data used to directly determine the weights and thresholds of the model given inputs and targets (desired outputs). It's performance estimates are, obviously, highly biased.

VALIDATION SUBSET: Nontraining design data used two ways:

1. Validation stopping: Training is stopped 
   when the error on the validation subset  
   increases for 6(default) consecutive epochs.
2. Ranking multiple designs with slightly 
   biased performance estimates that typically 
   differ by 
   a. number of hidden nodes  
   b. initial random weight assignments.

TESTING SUBSET: Once designs are ranked to be satisfactory, UNBIASED PERFORMANCE ESTIMATES are obtained using the test subset.

 If no unbiased performance estimate is 
 satisfactory, the process can be repeated with 
 the data repartitioned and/or design parameters 
 changed.

MATLAB yields separate confusion matrices for the three data division subsets in addition to the one for all of the data .

Hope this helps.

Greg

Answer 2

[Greg Heath]

This is a VERY POOR classification confusion matrix layout

 1. There should be three confusion matrices
   a. One 5x5 count confusion matrix
   b. Two 5x5 percent confusion matrices

 2. For the count confusion matrix
   a. The rows should indicate how many of 
      each target class are classified into 
      each output class
   b. The last column should contain the 
      total number of members of each 
      "T"arget class
   c. The last row should contain the total 
      number of members "A"ssigned to each 
      class

       1     2    3   4   | T
    ----------------------------    
   1| 6670   28  801  501 | 8000 
   2|   86 6476  762  676 | 8000
   3| 1234  625 5934  207 | 8000
   4|  558  804  127 6511 | 8000
    ----------------------------
   A| 8548 7933 7624 7895 | 3200

   3. For the row percent confusion matrix 
   each row is divided by the total in column 
   T and multiplied by 100. This indicates 
   what percent of each class is assigned to 
   each output. 

   4. For the column percent confusion matrix 
   each column is divided by the total in row
   A and multiplied by 100. This indicates what 
   percent of each output is caused by which 
   target class.

   5. THE MATLAB CONVENTION
      a. The count confusion is TRANSPOSED from
      the classical convention.
      b. It is not obvious what the percentages 
      mean.
      c. I figured it out some years ago. Then I 
      shook my head in bewilderment and tried to 
      forget it. It looks like I succeeded.

Hope this helps.

Good Luck.

Greg

[Swathi Ganesh]

Thank You Greg. This plot is generated by the neural network app in matlab. I did not come across row and count confusion matrices in Matlab. I am attaching another image which has 4 confusion matrix plots.