Saved in:
Bibliographic Details
Main Authors: Zhang, Yuanhang, Lin, Zhidi, Sun, Yiyong, Yin, Feng, Fritsche, Carsten
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2403.10123
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916306619465728
author Zhang, Yuanhang
Lin, Zhidi
Sun, Yiyong
Yin, Feng
Fritsche, Carsten
author_facet Zhang, Yuanhang
Lin, Zhidi
Sun, Yiyong
Yin, Feng
Fritsche, Carsten
contents Deep state-space models (DSSMs) have gained popularity in recent years due to their potent modeling capacity for dynamic systems. However, existing DSSM works are limited to single-task modeling, which requires retraining with historical task data upon revisiting a forepassed task. To address this limitation, we propose continual learning DSSMs (CLDSSMs), which are capable of adapting to evolving tasks without catastrophic forgetting. Our proposed CLDSSMs integrate mainstream regularization-based continual learning (CL) methods, ensuring efficient updates with constant computational and memory costs for modeling multiple dynamic systems. We also conduct a comprehensive cost analysis of each CL method applied to the respective CLDSSMs, and demonstrate the efficacy of CLDSSMs through experiments on real-world datasets. The results corroborate that while various competing CL methods exhibit different merits, the proposed CLDSSMs consistently outperform traditional DSSMs in terms of effectively addressing catastrophic forgetting, enabling swift and accurate parameter transfer to new tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2403_10123
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Regularization-Based Efficient Continual Learning in Deep State-Space Models
Zhang, Yuanhang
Lin, Zhidi
Sun, Yiyong
Yin, Feng
Fritsche, Carsten
Machine Learning
Deep state-space models (DSSMs) have gained popularity in recent years due to their potent modeling capacity for dynamic systems. However, existing DSSM works are limited to single-task modeling, which requires retraining with historical task data upon revisiting a forepassed task. To address this limitation, we propose continual learning DSSMs (CLDSSMs), which are capable of adapting to evolving tasks without catastrophic forgetting. Our proposed CLDSSMs integrate mainstream regularization-based continual learning (CL) methods, ensuring efficient updates with constant computational and memory costs for modeling multiple dynamic systems. We also conduct a comprehensive cost analysis of each CL method applied to the respective CLDSSMs, and demonstrate the efficacy of CLDSSMs through experiments on real-world datasets. The results corroborate that while various competing CL methods exhibit different merits, the proposed CLDSSMs consistently outperform traditional DSSMs in terms of effectively addressing catastrophic forgetting, enabling swift and accurate parameter transfer to new tasks.
title Regularization-Based Efficient Continual Learning in Deep State-Space Models
topic Machine Learning
url https://arxiv.org/abs/2403.10123