SiT: Symmetry-Invariant Transformers for Generalisation in Reinforcement Learning.
Matthias WeissenbacherRishabh AgarwalYoshinobu KawaharaPublished in: CoRR (2024)
Keyphrases
- reinforcement learning
- invariance properties
- reinforcement learning algorithms
- bilateral symmetry
- planar shapes
- function approximation
- symmetry detection
- state space
- markov decision processes
- affine transformation
- multi agent
- optimal policy
- learning process
- optimal control
- affine invariant
- model free
- translation invariant
- temporal difference learning
- quasi invariant
- geometric transformations
- learning problems
- learning algorithm
- robotic control
- invariant properties
- temporal difference
- planar objects
- reinforcement learning methods
- function approximators
- robot control
- reward function
- neural network
- supervised learning
- computer vision
- machine learning