deep learning has led to remarkable advances when applied to problems where the data distribution does not change over the course of learning. In stark contrast, biological neural networks continually adapt to ch