Enum ReplaySamplingStrategy

Namespace

AiDotNet.ContinualLearning.Memory

Assembly

AiDotNet.dll

Strategy for sampling during replay (training time).

public enum ReplaySamplingStrategy

Fields

PriorityBased = 2

Priority-based sampling using importance weights. Examples that caused high loss are sampled more frequently.

RecencyWeighted = 3

Recency-weighted sampling favors more recent examples. Based on the hypothesis that recent examples are more relevant.

TaskBalanced = 1

Task-balanced sampling ensures equal representation of each task. Prevents bias toward tasks with more stored examples.

Uniform = 0

Uniform random sampling from the buffer.