/genetic_deep_learning

This project aims to use genetic algorithms to boost the learning of DNN. Building and training a family of NN with same structure and hyperparameters from scratch but starting from different random weights. After a few epochs of training, the networks that perform better are chosen and crossover their weights in order to mating and produce the next generation.

Primary LanguageJupyter NotebookOtherNOASSERTION

Stargazers