Continual learning with hypernetworksDownload PDF

Published: 19 Oct 2020, Last Modified: 17 Sept 2023ICLR 2020 Conference Blind SubmissionReaders: Everyone
Original Pdf: pdf
Code: https://github.com/chrhenning/hypercl
Community Implementations: [![CatalyzeX](/images/catalyzex_icon.svg) 4 code implementations](https://www.catalyzex.com/paper/arxiv:1906.00695/code)
Abstract: Artificial neural networks suffer from catastrophic forgetting when they are sequentially trained on multiple tasks. To overcome this problem, we present a novel approach based on task-conditioned hypernetworks, i.e., networks that generate the weights of a target model based on task identity. Continual learning (CL) is less difficult for this class of models thanks to a simple key feature: instead of recalling the input-output relations of all previously seen data, task-conditioned hypernetworks only require rehearsing task-specific weight realizations, which can be maintained in memory using a simple regularizer. Besides achieving state-of-the-art performance on standard CL benchmarks, additional experiments on long task sequences reveal that task-conditioned hypernetworks display a very large capacity to retain previous memories. Notably, such long memory lifetimes are achieved in a compressive regime, when the number of trainable hypernetwork weights is comparable or smaller than target network size. We provide insight into the structure of low-dimensional task embedding spaces (the input space of the hypernetwork) and show that task-conditioned hypernetworks demonstrate transfer learning. Finally, forward information transfer is further supported by empirical results on a challenging CL benchmark based on the CIFAR-10/100 image datasets.
Keywords: Continual Learning, Catastrophic Forgetting, Meta Model, Hypernetwork
9 Replies

Loading