Improving Deep Neuroevolution via Deep Innovation Protection

12/29/2019
by   Sebastian Risi, et al.
10

Evolutionary-based optimization approaches have recently shown promising results in domains such as Atari and robot locomotion but less so in solving 3D tasks directly from pixels. This paper presents a method called Deep Innovation Protection (DIP) that allows training complex world models end-to-end for such 3D environments. The main idea behind the approach is to employ multiobjective optimization to temporally reduce the selection pressure on specific components in a world model, allowing other components to adapt. We investigate the emergent representations of these evolved networks, which learn a model of the world without the need for a specific forward-prediction loss.

READ FULL TEXT

page 4

page 5

page 6

page 7

research
04/08/2019

Evolved embodied phase coordination enables robust quadruped robot locomotion

Overcoming robotics challenges in the real world requires resilient cont...
research
04/28/2019

Deep Neuroevolution of Recurrent and Discrete World Models

Neural architectures inspired by our own human cognitive system, such as...
research
03/16/2017

End-to-End Learning for Structured Prediction Energy Networks

Structured Prediction Energy Networks (SPENs) are a simple, yet expressi...
research
03/26/2021

Quantitative Assessment of Solution Innovation in Engineering Education

The article discusses the quantitative assessment approach to the innova...
research
07/05/2017

Data-Driven Sparse Structure Selection for Deep Neural Networks

Deep convolutional neural networks have liberated its extraordinary powe...
research
06/09/2021

Self-Improved Retrosynthetic Planning

Retrosynthetic planning is a fundamental problem in chemistry for findin...

Please sign up or login with your details

Forgot password? Click here to reset