How to interpret the result files

The results of an experiment is stored in the mmm folder generated. In this folder, you will find:

1. Archive files.

• E.g. archive_1000.dat, which is the file generated after 1000 generations containing the fitness or prediction scores of all 1000 images generated.

• You will find two columns of numbers, which look something like below:

   7 0.0727381
   8 0.0317073
   9 0.0296667
   10 0.0185064
  

• 10 here is the category #10 in 1000 ImageNet categories. (This column ranges from 0 to 999)

• 0.0185064 is the prediction score of best image evolved in that category. In this case, the image has the confidence of ~1.85%.

2. Images folders.

• E.g. map_gen_1000, which is the folder generated after 1000 generations containing the final images evolved. These images correspond to the entries in the archive_1000.dat.
• A file would look like: map_1000_9_0.32.png. This means that:
  • This file is generated after 1000 generations
  • This image is the best (highest score) in the category #9 out of 1000 ImageNet categories.
  • This image is given 32% prediction score by the DNN model.