Learning to Generate Genotypes with Neural Networks

TL;DR

This paper explores using neural networks to generate mutation distributions in evolutionary algorithms, demonstrating improved performance on complex discrete problems like MaxSat and HIFF.

Contribution

It introduces a novel approach where neural networks are used to learn mutation distributions, offering advantages over traditional methods in evolutionary computation.

Findings

Neural network-based mutation distributions outperform traditional methods on MaxSat.

The approach effectively solves complex discrete optimization problems.

Neural models like Denoising Autoencoders and NADE are effective in this context.

Abstract

Neural networks and evolutionary computation have a rich intertwined history. They most commonly appear together when an evolutionary algorithm optimises the parameters and topology of a neural network for reinforcement learning problems, or when a neural network is applied as a surrogate fitness function to aid the evolutionary optimisation of expensive fitness functions. In this paper we take a different approach, asking the question of whether a neural network can be used to provide a mutation distribution for an evolutionary algorithm, and what advantages this approach may offer? Two modern neural network models are investigated, a Denoising Autoencoder modified to produce stochastic outputs and the Neural Autoregressive Distribution Estimator. Results show that the neural network approach to learning genotypes is able to solve many difficult discrete problems, such as MaxSat and HIFF, and regularly outperforms other evolutionary techniques.