REWRITE GENETIC CNN BY TORCH

Rewrite for torch with following paper: Lingxi Xie and Alan L. Yuille Genetic CNN

Class format flow

Visualize Class format flow

Stage (complete)

Example - num_nodes = 5 - binary_code = 1100000001 - separated_connections = 1-10-000-0001

Stages (complete)

Example - num_stages = (6, 9, 8) - binary_code_stages_model = {'S_1': '1-00-010-0110-00100', 'S_2': '0-00-000-1001-01101-001110-0101101-00001110', 'S_3': '1-01-001-1100-10011-011100-1100110'}

Stages			Stage
stages	binary_code_stages_model		binary_code	num_nodes
	index	separated connections
0	S_1	1-00-010-0110-00100	100010011000100	6
1	S_2	0-00-000-1001-01101-001110-0101101-00001110	000000100101101001110010110100001110	9
2	S_3	1-01-001-1100-10011-011100-1100110	1010011100100110111001100110	8

Individual (complete)

run cells: import, class Individual, Train & test model for Individual

Training on single Individual with MNIST run 5 epochs

Population & Genetic Algorithm (complete)

run cells: import, class Population, test

See example output in files test1.log, test2.log

ENVIRONMENT REQUIREMENTS

• pytorch >= 1.9