Regularization-Based Efficient Continual Learning in Deep State-Space Models
This work addresses the problem of efficient continual learning for dynamic systems modeling, which is incremental as it applies existing regularization-based methods to DSSMs.
The paper tackles the problem of catastrophic forgetting in deep state-space models (DSSMs) when adapting to multiple tasks, proposing continual learning DSSMs (CLDSSMs) that integrate regularization-based methods to enable efficient updates with constant computational and memory costs, and experiments on real-world datasets show CLDSSMs consistently outperform traditional DSSMs in addressing forgetting and enabling swift parameter transfer.
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.