Prioritized Experience Replay

AI and Machine Learning · Reinforcement Learning · 2015 · also: PER · prioritized-experience-replay.yaml

Draw a transition in proportion to the size of its last TD error, so the buffer keeps returning the surprising ones and stops returning the solved ones.

Colour is the family; a dashed line is the second member of it.

Drag to pan · scroll to zoom · click a node to open it
G experience-replay Experience Replay importance-sampling Importance Sampling prioritized-experience-replay Prioritized Experience Replay prioritized-experience-replay->experience-replay uniform sampling spends most updates on transitions with no error left prioritized-experience-replay->importance-sampling drawing by error skews the expectation, and the weight puts it back rainbow Rainbow rainbow->prioritized-experience-replay no single ablation costs more than removing it

This node

Referenced by

References