Attention or memory? Neurointerpretable agents in space and time

07/09/2020
by   Lennart Bramlage, et al.
0

In neuroscience, attention has been shown to bidirectionally interact with reinforcement learning (RL) processes. This interaction is thought to support dimensionality reduction of task representations, restricting computations to relevant features. However, it remains unclear whether these properties can translate into real algorithmic advantages for artificial agents, especially in dynamic environments. We design a model incorporating a self-attention mechanism that implements task-state representations in semantic feature-space, and test it on a battery of Atari games. To evaluate the agent's selective properties, we add a large volume of task-irrelevant features to observations. In line with neuroscience predictions, self-attention leads to increased robustness to noise compared to benchmark models. Strikingly, this self-attention mechanism is general enough, such that it can be naturally extended to implement a transient working-memory, able to solve a partially observable maze task. Lastly, we highlight the predictive quality of attended stimuli. Because we use semantic observations, we can uncover not only which features the agent elects to base decisions on, but also how it chooses to compile more complex, relational features from simpler ones. These results formally illustrate the benefits of attention in deep RL and provide evidence for the interpretability of self-attention mechanisms.

READ FULL TEXT

page 2

page 3

page 4

page 8

page 9

page 10

page 13

page 16

research
03/18/2020

Neuroevolution of Self-Interpretable Agents

Inattentional blindness is the psychological phenomenon that causes one ...
research
06/14/2022

Transformers are Meta-Reinforcement Learners

The transformer architecture and variants presented remarkable success a...
research
10/24/2022

Composition, Attention, or Both?

In this paper, we propose a novel architecture called Composition Attent...
research
11/08/2021

E(2) Equivariant Self-Attention for Radio Astronomy

In this work we introduce group-equivariant self-attention models to add...
research
09/15/2022

Towards self-attention based visual navigation in the real world

Vision guided navigation requires processing complex visual information ...
research
11/17/2019

Working Memory Graphs

Transformers have increasingly outperformed gated RNNs in obtaining new ...

Please sign up or login with your details

Forgot password? Click here to reset