Sparsity and heterogeneous dropout for continual learning in the null space of neural activations

Thumbnail Image

Files

abbasi22a.pdf (1.3 MB)

Links to Files

https://proceedings.mlr.press/v199/abbasi22a

Permanent Link

http://hdl.handle.net/11603/26318

Collections

UMBC Computer Science and Electrical Engineering Department
UMBC Faculty Collection

Author/Creator

Author/Creator ORCID

Date

2022

Type of Work

conference papers and proceedings
preprints
Text

Department

Program

Citation of Original Publication

Abbasi, Ali, Parsa Nooralinejad, Vladimir Braverman, Hamed Pirsiavash, and Soheil Kolouri. “Sparsity and Heterogeneous Dropout for Continual Learning in the Null Space of Neural Activations.” In Proceedings of The 1st Conference on Lifelong Learning Agents, 617–28. PMLR, 2022. https://proceedings.mlr.press/v199/abbasi22a.html.

Rights

This item is likely protected under Title 17 of the U.S. Copyright Law. Unless on a Creative Commons license, for uses protected by Copyright Law, contact the copyright holder or the author.

Subjects

Abstract

Continual/lifelong learning from a non-stationary input data stream is a cornerstone of intelligence. Despite their phenomenal performance in a wide variety of applications, deep neural networks are prone to forgetting their previously learned information upon learning new ones. This phenomenon is called “catastrophic forgetting” and is deeply rooted in the stability-plasticity dilemma. Overcoming catastrophic forgetting in deep neural networks has become an active field of research in recent years. In particular, gradient projection-based methods have recently shown exceptional performance at overcoming catastrophic forgetting. This paper proposes two biologically-inspired mechanisms based on sparsity and heterogeneous dropout that significantly increase a continual learner’s performance over a long sequence of tasks. Our proposed approach builds on the Gradient Projection Memory (GPM) framework. We leverage k-winner activations in each layer of a neural network to enforce layer-wise sparse activations for each task, together with a between-task heterogeneous dropout that encourages the network to use non-overlapping activation patterns between different tasks. In addition, we introduce two new benchmarks for continual learning under distributional shift, namely Continual Swiss Roll and ImageNet SuperDog-40. Lastly, we provide an in-depth analysis of our proposed method and demonstrate a significant performance boost on various benchmark continual learning problems.