I wonder if there is research, patents, or libraries using Genetic algorithms (GA) to improve Neural Networks. I don't find anything in the subject. For example:
- use GA to find better parameters in a NN. So the chromosome will be [learning rate, activation function, layers number, layers size, dropout factor] and the fit function minimize computational cost to reach NN 95% accuracy.
- use GA to mix your NN input data and generate new data to adjust.
- use GA to mix several small NN, different types, and find the perfect mix for better predictions.