Recurrent World Models Facilitate Policy Evolution

09/04/2018
by   David Ha, et al.
0

A generative recurrent neural network is quickly trained in an unsupervised manner to model popular reinforcement learning environments through compressed spatio-temporal representations. The world model's extracted features are fed into compact and simple policies trained by evolution, achieving state of the art results in various environments. We also train our agent entirely inside of an environment generated by its own internal world model, and transfer this policy back into the actual environment. Interactive version of paper at https://worldmodels.github.io

READ FULL TEXT
research
03/27/2018

World Models

We explore building generative neural network models of popular reinforc...
research
06/17/2019

Iterative Model-Based Reinforcement Learning Using Simulations in the Differentiable Neural Computer

We propose a lifelong learning architecture, the Neural Computer Agent (...
research
10/27/2022

Meta-Reinforcement Learning Using Model Parameters

In meta-reinforcement learning, an agent is trained in multiple differen...
research
03/06/2019

Using World Models for Pseudo-Rehearsal in Continual Learning

The utility of learning a dynamics/world model of the environment in rei...
research
07/15/2021

Adaptable Agent Populations via a Generative Model of Policies

In the natural world, life has found innumerable ways to survive and oft...
research
10/05/2020

Mastering Atari with Discrete World Models

Intelligent agents need to generalize from past experience to achieve go...
research
09/17/2021

Dropout's Dream Land: Generalization from Learned Simulators to Reality

A World Model is a generative model used to simulate an environment. Wor...

Please sign up or login with your details

Forgot password? Click here to reset