Attention-based Experience Replay Framework for Continual Learning of Agnostic Time Series Forecasting Models
Summary
This paper introduces a novel continual learning framework for time series forecasting that uses attention-guided experience replay to enable models to adapt to new data distributions while avoiding catastrophic forgetting, evaluated on benchmarks and real-world piezometric data.
View Cached Full Text
Cached at: 07/24/26, 05:02 AM
# Attention-based Experience Replay Framework for Continual Learning of Agnostic Time Series Forecasting Models Source: [https://arxiv.org/abs/2607.20493](https://arxiv.org/abs/2607.20493) [View PDF](https://arxiv.org/pdf/2607.20493) > Abstract:Deep learning has led to remarkable progress in artificial intelligence, particularly in robotics, imaging and sound processing\. However, a major limitation of neural networks remains their strong dependence on large and stationary datasets\. In many real\-world applications, these conditions are rarely met due to evolving and dynamic environments where data distributions change over time\. Continual learning aims to address this challenge by developing models capable of adapting incrementally while maintaining a balance between stability and plasticity under computational constraints\. In this work, we introduce a novel framework for continual time series forecasting, designed to extend existing static forecasting models commonly used in the literature by incorporating an Experience Replay strategy guided by Attention mechanisms\. This approach allows the model to adapt dynamically to new contexts while preserving prior knowledge, effectively mitigating catastrophic forgetting\. The framework is evaluated on standard forecasting benchmarks as well as on a piezometric dataset exhibiting diverse temporal behaviors\. Results show that our approach effectively increases or maintains predictive performance over time while reducing retraining costs and data requirements, thus facilitating the deployment of forecasting models in dynamic and real\-world settings\. ## Submission history From: Quentin Besnard \[[view email](https://arxiv.org/show-email/877aa9f2/2607.20493)\] \[via CCSD proxy\] **\[v1\]**Fri, 12 Jun 2026 11:44:05 UTC \(925 KB\)
Similar Articles
Rethinking Transfer in Continual Learning: A Replay-Based Realisation
This paper introduces a framework for when transfer should be expected in continual learning and proposes Transfer-Selective Replay (TSR), which selects replay data predicted to benefit the incoming task rather than indiscriminately replaying past examples. TSR improves forward transfer while maintaining stability, outperforming existing replay baselines.
Towards Continuous Power Forecasting: Practical Continual Learning for Real-World Energy Systems in Nonstationary Time Series
This paper proposes Continuous Power Forecasting, treating power forecasting as a continual learning problem to handle nonstationary conditions. It evaluates six CL approaches on real-world datasets, showing benefits in adaptation and mitigating catastrophic forgetting.
Continual Learning with Elastic Regularization and Synthetic Replay for Federated MLLM Fine-Tuning
Proposes FedCMM, a framework for federated continual learning of multimodal LLMs that uses modality-aware elastic weight consolidation, local generative replay, and task-similarity-aware gradient aggregation to mitigate catastrophic forgetting.
Freshness-Aware Prioritized Experience Replay for LLM/VLM Reinforcement Learning
FreshPER introduces a freshness-aware prioritized experience replay method for LLM/VLM reinforcement learning that addresses the 'priority staleness' problem by applying exponential age decay to stored priorities, enabling off-policy reuse of trajectories. Evaluated on eight agentic, reasoning, and math tasks, FreshPER significantly outperforms on-policy baselines with gains up to +367% on Sokoban.
Overcoming Catastrophic Forgetting in Visual Continual Learning with Reinforcement Fine-Tuning
This paper introduces Retention-aware Policy Optimization (RaPO) to mitigate catastrophic forgetting in visual continual learning using reinforcement fine-tuning. RaPO uses trajectory-level reward shaping and cross-task advantage normalization to close the gap between reinforcement and supervised fine-tuning in class- and domain-incremental learning.