Behavioral Experiments for Understanding Catastrophic Forgetting

TL;DR

This paper applies behavioral experimental methods from psychology to study catastrophic forgetting in neural networks, providing new insights and demonstrating a behavior-first approach to understanding neural phenomena.

Contribution

It introduces a behavioral experiment framework for neural networks, offering novel insights into catastrophic forgetting and proposing an alternative investigative approach.

Findings

Controlled experiments reveal new behaviors of catastrophic forgetting.

Behavioral methods provide valuable insights into neural network phenomena.

Proposes a behavior-first approach as an alternative to traditional analysis.

Abstract

In this paper we explore whether the fundamental tool of experimental psychology, the behavioral experiment, has the power to generate insight not only into humans and animals, but artificial systems too. We apply the techniques of experimental psychology to investigating catastrophic forgetting in neural networks. We present a series of controlled experiments with two-layer ReLU networks, and exploratory results revealing a new understanding of the behavior of catastrophic forgetting. Alongside our empirical findings, we demonstrate an alternative, behavior-first approach to investigating neural network phenomena.