LACE: Loss-Adaptive Capacity Expansion for Continual Learning

TL;DR

LACE is an online method that adaptively expands a model's capacity during training based on loss signals, improving continual learning efficiency without prior knowledge of data complexity.

Contribution

LACE introduces a loss-based capacity expansion mechanism that dynamically adjusts model size during training, eliminating the need for pre-specified capacity in continual learning.

Findings

LACE triggers expansions precisely at domain boundaries with 100% boundary detection accuracy.

It matches large fixed-capacity model accuracy while using fewer parameters.

Removing all adapters causes only a 3% drop in accuracy, showing their importance.

Abstract

Fixed representational capacity is a fundamental constraint in continual learning: practitioners must guess an appropriate model width before training, without knowing how many distinct concepts the data contains. We propose LACE (Loss-Adaptive Capacity Expansion), a simple online mechanism that expands a model's representational capacity during training by monitoring its own loss signal. When sustained loss deviation exceeds a threshold - indicating that the current capacity is insufficient for newly encountered data - LACE adds new dimensions to the projection layer and trains them jointly with existing parameters. Across synthetic and real-data experiments, LACE triggers expansions exclusively at domain boundaries (100% boundary precision, zero false positives), matches the accuracy of a large fixed-capacity model while starting from a fraction of its dimensions, and produces adapter dimensions that are collectively critical to performance (3% accuracy drop when all adapters removed). We further demonstrate unsupervised domain separation in GPT-2 activations via layer-wise clustering, showing a U-shaped separability curve across layers that motivates adaptive capacity allocation in deep networks. LACE requires no labels, no replay buffers, and no external controllers, making it suitable for on-device continual learning under resource constraints.