Heads collapse, features stay: Why Replay needs big buffers

Giulia Lanzillotta, Damiano Meier, Thomas Hofmann

Keywords: continual larning, neural collapse, deep learning

TL;DR: We analyze continual learning in the long-training limit, showing via Neural Collapse that replay preserves feature separability but causes head–feature misalignment, explaining why deep forgetting is mitigated while shallow forgetting persists.

Abstract:

A persistent paradox in continual learning (CL) is that neural networks often retain linearly separable representations of past tasks even when their output predictions fail. We formalize this distinction as the gap between *deep* (feature-space) and *shallow* (classifier-level) forgetting. We reveal a critical asymmetry in Experience Replay: while minimal buffers successfully anchor feature geometry and prevent deep forgetting, mitigating shallow forgetting typically requires substantially larger buffer capacities.
To explain this, we extend the Neural Collapse framework to the sequential setting. We characterize deep forgetting as a geometric drift toward out-of-distribution subspaces and prove that any non-zero replay fraction asymptotically guarantees the retention of linear separability. Conversely, we identify that the ``strong collapse'' induced by small buffers leads to rank-deficient covariances and inflated class means, effectively blinding the classifier to true population boundaries. By unifying CL with out-of-distribution detection, our work challenges the prevailing reliance on large buffers, suggesting that explicitly correcting these statistical artifacts could unlock robust performance with minimal replay.

Primary Area: transfer learning, meta learning, and lifelong learning

Submission Number: 7768