Evolving Reusable Neural Modules

Crawling Out of the Simulation: Evolving Real Robot Morphologies Using Cheap, Reusable Modules

A current issue in evolutionary robotics involves the coevolution of robot controllers and body morphologies built from modular parts. As part of ongoing research, a model for the evolution of the morphologies and neural network controllers of robots is described. Several robots are evolved for locomotion in simulation built from modules representing cheap, preexisting parts and one is physical...

Cooperative Co-Evolving Neural Networks for Robosoccer Simulation

Evolving Neuro - Modules and their Interfacesto Control

An evolutionary algorithm for the creation of recurrent network structures is presented. The aim is to develop neural networks controlling the behaviour of miniature robots. Two neuro-modules are created separately using this evolutionary approach. The rst neuro-module gives the agents the ability to move within an environment without colliding with obstacles. The second neuro-module provides t...

Smaller Architecture Search Spaces

To simplify neural architecture creation, AutoML is gaining traction from evolutionary algorithms to reinforcement learning or simple search in a constrained space of neural modules. A big issues is its computational cost: the size of the search space can easily go above ̃10 candidates for a 10-layer network and the cost of evaluating a single candidate is high even if it’s not fully trained. I...

Evolutionary Architecture Search For Deep Multitask Networks

Multitask learning, i.e. learning several tasks at once with the same neural network, can improve performance in each of the tasks. Designing deep neural network architectures for multitask learning is a challenge: There are many ways to tie the tasks together, and the design choices matter. The size and complexity of this problem exceeds human design ability, making it a compelling domain for ...