Learn to Bind and Grow Neural Structures

TL;DR

This paper introduces a novel framework for task-incremental learning that adaptively binds or expands neural network structures using Bayesian optimization, balancing performance and model size.

Contribution

It proposes an interpretable parameterization of multi-task architectures allowing globally optimal solutions for incremental learning via Bayesian optimization.

Findings

Performs comparably with existing expansion methods on benchmarks.

Can compute multiple optimal architectures with different performance-size trade-offs.

Flexible framework for continual learning with structure adaptation.

Abstract

Task-incremental learning involves the challenging problem of learning new tasks continually, without forgetting past knowledge. Many approaches address the problem by expanding the structure of a shared neural network as tasks arrive, but struggle to grow optimally, without losing past knowledge. We present a new framework, Learn to Bind and Grow, which learns a neural architecture for a new task incrementally, either by binding with layers of a similar task or by expanding layers which are more likely to conflict between tasks. Central to our approach is a novel, interpretable, parameterization of the shared, multi-task architecture space, which then enables computing globally optimal architectures using Bayesian optimization. Experiments on continual learning benchmarks show that our framework performs comparably with earlier expansion based approaches and is able to flexibly compute multiple optimal solutions with performance-size trade-offs.