Bootstrap Methods in RL

Dropout as Bootstrap Approximation

Technique using dropout during inference as an efficient approximation of bootstrap to quickly estimate uncertainty without training multiple models.

← Back