Skip to content

mikhliuk-k/neural_network_compression_picture

Repository files navigation

Recycling network - compression picture

The dependence of the number of iterations of the compression ratio

The following input parameters were set to determine this relationship:

  • 256x256 size image;
  • Squares of the same image size 4x4;
  • The maximum permissible errors - in 1000.

Changing the compression ratio is achieved by varying the number of hidden layer neurons.

Compression coefficient Number of iteration
4,20089 40
3,15069 17
2,52056 8
2,10047 6
1,80041 5

The dependence of the number of training iterations, for a different size image

The following input parameters were set to determine this relationship:

  • 256x256 sized image;
  • squares of the same image size 4x4;
  • The number of neurons in the hidden layer - 28
  • maximum permissible errors - 500.

Use a single image in different sizes.

Image Size Number of iterations
32 x 32 3
64 x 64 10
128 x 128 15
256 x 256 17

The dependence of the number of iterations of the error

The following input parameters were set to determine this relationship:

  • 256x256 sized image;
  • squares of the same image size 4x4;
  • The number of neurons in the hidden layer - 20;
Error Number of iterations
2086,42 1
1739,09 2
1285,45 4
1021,98 6
857,13 8
733,34 10
634,46 12
556,18 14
495,15 16

The dependence of the number of iterations of step

The following input parameters were set to determine this relationship:

  • 256x256 sized image;
  • squares of the same image size 4x4;
  • The number of neurons in the hidden layer - 32;
  • maximum permissible errors - 500.
Step training Number of iterations
0.0001 53
0.001 8
0.003 3
0.005 2
0.01 2

Example

demo

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published