Entropy-based Stability-Plasticity for Lifelong Learning

TL;DR

This paper introduces an entropy-based method to dynamically balance stability and plasticity in neural networks, improving lifelong learning by reducing interference and enabling layer freezing for faster training.

Contribution

The proposed Entropy-based Stability-Plasticity (ESP) method adaptively adjusts layer plasticity using entropy criteria, addressing the stability-plasticity dilemma in lifelong learning.

Findings

Effective in natural language and vision tasks

Reduces interference and leverages prior knowledge

Enables layer freezing to speed up training

Abstract

The ability to continuously learn remains elusive for deep learning models. Unlike humans, models cannot accumulate knowledge in their weights when learning new tasks, mainly due to an excess of plasticity and the low incentive to reuse weights when training a new task. To address the stability-plasticity dilemma in neural networks, we propose a novel method called Entropy-based Stability-Plasticity (ESP). Our approach can decide dynamically how much each model layer should be modified via a plasticity factor. We incorporate branch layers and an entropy-based criterion into the model to find such factor. Our experiments in the domains of natural language and vision show the effectiveness of our approach in leveraging prior knowledge by reducing interference. Also, in some cases, it is possible to freeze layers during training leading to speed up in training.