Learning Dynamics and Pathologies
How learning evolves, why it stalls or degrades, and what enables it to scale.
How learning evolves, why it stalls or degrades, and what enables it to scale.
How new experience and feedback improve an agent’s knowledge, strategies, and skills.
How agents direct their own learning to acquire reliable, reusable robot skills.
Useful visual diversity depends on both the transformation and the way it enters training.
Data, the critic and the timing of updates all shape an agent's learning capacity.
Static sparsity offers a surprisingly simple way to improve scaling in deep RL.