Maslow's Hammer for Catastrophic Forgetting: Node Re-Use vs Node Activation

TL;DR

This paper investigates catastrophic forgetting in neural networks, revealing a trade-off between node activation and re-use that explains why forgetting peaks at intermediate task similarities, and reinterprets existing methods accordingly.

Contribution

It introduces the Maslow's hammer hypothesis, explaining the non-monotonic forgetting pattern and analyzes the effectiveness of interventions based on this trade-off.

Findings

Forgetting peaks at intermediate task similarity regimes.

Trade-off between node activation and re-use explains forgetting behavior.

Reinterpretation of algorithms based on the trade-off.

Abstract

Continual learning - learning new tasks in sequence while maintaining performance on old tasks - remains particularly challenging for artificial neural networks. Surprisingly, the amount of forgetting does not increase with the dissimilarity between the learned tasks, but appears to be worst in an intermediate similarity regime. In this paper we theoretically analyse both a synthetic teacher-student framework and a real data setup to provide an explanation of this phenomenon that we name Maslow's hammer hypothesis. Our analysis reveals the presence of a trade-off between node activation and node re-use that results in worst forgetting in the intermediate regime. Using this understanding we reinterpret popular algorithmic interventions for catastrophic interference in terms of this trade-off, and identify the regimes in which they are most effective.