The Sensory Neuron as a Transformer: Permutation-Invariant Neural Networks for Reinforcement Learning

Permutation-Invariant Neural Networks for Reinforcement Learning

--> This page requires Javascript. Please enable it for https://attentionagent.github.io/ Examples of permutation-invariant reinforcement learning agents In this work, we investigate the properties of RL agents that treat their observations as an arbitrarily ordered, variable-length list of sensory inputs. Here, we partition the visual input from CarRacing (Left) and Atari Pong (right) into a 2D grid of small patches, and shuffled their ordering. Each sensory neuron in the system receives a stream of visual input at a particular permuted patch location, and through coordination, must complete

related reading

• Toy Models of Superpositiontransformer-circuits.pub
• [1810.00825] Set Transformer: A Framework for Attention-based Permutation-Invariant Neural Networksarxiv.org
• Deep Reinforcement Learning from Human Preferencesproceedings.neurips.cc
• Set Transformer: A Framework for Attention-basedPermutation-Invariant Neural Networksarxiv.org
• pdfopenreview.net
• How our data shaped neural architecture discovery, and how automation can reshape the future | Core Automationcoreauto.com
• Deep Reinforcement Learning: Pong from Pixelskarpathy.github.io
• 2106.01345arxiv.org
• Multimodal Neurons in Artificial Neural Networksdistill.pub
• Feature-wise transformationsdistill.pub
• Dynamic Inference with Neural Interpretersarxiv.org
• Naturally Occurring Equivariance in Neural Networksdistill.pub