Neurogenetic Programming Framework for Explainable Reinforcement Learning

TL;DR

This paper introduces a neurogenetic programming framework that combines evolutionary algorithms with neural networks to produce explainable and high-performing reinforcement learning solutions, integrating expert knowledge into the process.

Contribution

It presents a novel hybrid neuro-genetic approach that leverages evolutionary methods for neural network training, enhancing explainability and performance in reinforcement learning tasks.

Findings

Effective in OpenAI Gym environments

Provides interpretable solutions

Allows injection of expert knowledge

Abstract

Automatic programming, the task of generating computer programs compliant with a specification without a human developer, is usually tackled either via genetic programming methods based on mutation and recombination of programs, or via neural language models. We propose a novel method that combines both approaches using a concept of a virtual neuro-genetic programmer: using evolutionary methods as an alternative to gradient descent for neural network training}, or scrum team. We demonstrate its ability to provide performant and explainable solutions for various OpenAI Gym tasks, as well as inject expert knowledge into the otherwise data-driven search for solutions.