Name: Continual Learning Lab
Author: MySimulator

Training

Learning rate: 0.15

Enable replay buffer (mix in old Task-A samples while training Task B)

Task A and Task B are two different linear classification boundaries over the same 2D input space — a minimal stand-in for "two skills sharing one small network."

Accuracy

Task A accuracy—

Task B accuracy—

Stagenot trained

Decision boundary (Task A points = circles, Task B points = squares)