Skip to main content

Project Synopsis

Problem Being Addressed:

Deep learning convolutional neural networks require extensive manual trial and error with expensive resources to find an optimal parameter set.


Proposed Solution:

Implement a genetic algorithm to quickly find optimal parameters & hyperparameters for deep learning convolutional neural networks.

Our team implemented a genetic algorithm and specific techniques for different stages such as cross over, mutation, fitness function, & more and integrated this with Deep Learning to produce optimal CNN models for a given dataset such that the new architecture and parameters (example - filters, activation, optimizer, pool size etc.) will result in optimal accuracy and loss.