Abstract:Deep artificial neural networks (DNNs) are typically trained via gradient-based learning algorithms, namely backpropagation. Evolution strategies (ES) can rival backprop-based algorithms such as Q-learning and policy gradients on challenging deep reinforcement learning (RL) problems. However, ES can be considered a gradient-based algorithm because it performs stochastic gradient descent via an operation similar to a finite-difference approximation of the gradient. That raises the question of whether non-gradient-based evolutionary algorithms can work at DNN scales. Here we demonstrate they can: we evolve the weights of a DNN with a simple, gradient-free, population-based genetic algorithm (GA) and it performs well on hard deep RL problems, including Atari and humanoid locomotion. The Deep GA successfully evolves networks with over four million free parameters, the largest neural networks ever evolved with a traditional evolutionary algorithm. These results (1) expand our sense of the scale at which GAs can operate, (2) suggest intriguingly that in some cases following the gradient is not the best choice for optimizing performance, and (3) make immediately available the multitude of neuroevolution techniques that improve performance. We demonstrate the latter by showing that combining DNNs with novelty search, which encourages exploration on tasks with deceptive or sparse reward functions, can solve a high-dimensional problem on which reward-maximizing algorithms (e.g.\ DQN, A3C, ES, and the GA) fail. Additionally, the Deep GA is faster than ES, A3C, and DQN (it can train Atari in ${\raise.17ex\hbox{$\scriptstyle\sim$}}$4 hours on one desktop or ${\raise.17ex\hbox{$\scriptstyle\sim$}}$1 hour distributed on 720 cores), and enables a state-of-the-art, up to 10,000-fold compact encoding technique.

Is it conceivable that neurogenesis, neural Darwinism, and species evolution could all serve as inspiration for the creation of evolutionary deep neural networks?

Neuroevolution in Deep Neural Networks: Current Trends and Future Challenges

Towards Self-Assembling Artificial Neural Networks through Neural Developmental Programs

Impacts of Darwinian Evolution on Pre-trained Deep Neural Networks

A Survey on Evolutionary Construction of Deep Neural Networks.

Generative NeuroEvolution for Deep Learning

A Gradient-Guided Evolutionary Approach to Training Deep Neural Networks

Guest Editorial Evolutionary Neural Architecture Search

Evolutionary Shallowing Deep Neural Networks at Block Levels

Evolutionary neural networks for deep learning: a review

Deep Neuroevolution: Genetic Algorithms Are a Competitive Alternative for Training Deep Neural Networks for Reinforcement Learning

Brain-inspired Evolutionary Architectures for Spiking Neural Networks

A Survey on Evolutionary Neural Architecture Search

Evolutionary Neural Network Architecture Search

Biologically Inspired Structure Learning with Reverse Knowledge Distillation for Spiking Neural Networks

A Self-adaptive neuroevolution approach to constructing Deep Neural Network architectures across different types

Deep Connectomics Networks: Neural Network Architectures Inspired by Neuronal Networks

Neural Networks from Biological to Artificial and Vice Versa

Towards New Generation, Biologically Plausible Deep Neural Network Learning

An Artificial Neural Network Functionalized by Evolution