Skip to content
프로젝트

harvitronix/neural-network-genetic-algorithm

GitHub에서 보기

harvitronix/neural-network-genetic-algorithm has 702 GitHub stars and 230 forks. GitHot measured 0 stars in 24 hours and 0 in 7 days. Last checked 2026-11-21 09:30 UTC. Last push 2018-04-25.

이 수치를 측정하는 방법

측정된 스타 증가

in 24h
0
in 7 days
0
in 15 days
기록 수집 중
in 1 month
기록 수집 중

서로 겹치는 기간의 순증가량이며 일별 합계가 아닙니다. 측정값이 없으면 비워 둡니다. 스타는 관심도를 나타내며 품질을 보장하지 않습니다.

소개

Evolving a neural network with a genetic algorithm.

머신러닝

README.md

Original

Evolve a neural network with a genetic algorithm

This is an example of how we can use a genetic algorithm in an attempt to find the optimal network parameters for classification tasks.

It's currently limited to only MLPs (ie. fully connected networks) and uses the Keras library to build, train and validate.

On the easy MNIST dataset, we are able to quickly find a network that reaches > 98% accuracy. On the more challenging CIFAR10 dataset, we get to 56% after 10 generations (with population 20).

For more, see this blog post: https://medium.com/@harvitronix/lets-evolve-a-neural-network-with-a-genetic-algorithm-code-included-8809bece164

For a more robust implementation that you can use in your projects, take a look at Jan Liphardt's implementation, DeepEvolve.

To run

To run the brute force algorithm:

python3 brute.py

To run the genetic algorithm:

python3 main.py

You can set your network parameter choices by editing each of those files first. You can also choose whether to use the MNIST or CIFAR10 datasets. Simply set dataset to either mnist or cifar10.

Credits

The genetic algorithm code is based on the code from this excellent blog post: https://lethain.com/genetic-algorithms-cool-name-damn-simple/

Contributing

Have an optimization, idea, suggestion, bug report? Pull requests greatly appreciated!

License

MIT