Lifelong Learning with Dynamically Expandable Networks

TL;DR

This paper introduces a dynamically expandable neural network architecture for lifelong learning that adapts its capacity over time, sharing knowledge efficiently among tasks and preventing semantic drift, while outperforming existing methods.

Contribution

The paper presents a novel network architecture, DEN, that dynamically expands and manages capacity for lifelong learning, improving performance and parameter efficiency.

Findings

Outperforms existing lifelong learning methods

Achieves comparable performance to batch models with fewer parameters

Fine-tuned network on all tasks surpasses batch models

Abstract

We propose a novel deep network architecture for lifelong learning which we refer to as Dynamically Expandable Network (DEN), that can dynamically decide its network capacity as it trains on a sequence of tasks, to learn a compact overlapping knowledge sharing structure among tasks. DEN is efficiently trained in an online manner by performing selective retraining, dynamically expands network capacity upon arrival of each task with only the necessary number of units, and effectively prevents semantic drift by splitting/duplicating units and timestamping them. We validate DEN on multiple public datasets under lifelong learning scenarios, on which it not only significantly outperforms existing lifelong learning methods for deep networks, but also achieves the same level of performance as the batch counterparts with substantially fewer number of parameters. Further, the obtained network fine-tuned on all tasks obtained significantly better performance over the batch models, which shows that it can be used to estimate the optimal network structure even when all tasks are available in the first place.