ResCP: Reservoir Conformal Prediction for Time Series Forecasting

2025-10-06

AI Summary

Conformal prediction for time series forecasting faces challenges under distribution shift—including difficulty in constructing valid prediction intervals, poor reliability with limited data, and high retraining costs. Method: This paper proposes a training-free conformal prediction method that integrates reservoir computing into the conformal framework for the first time. It constructs local, time-aware conformity scores via dynamically reweighted residuals, enabling lightweight modeling of temporal dependencies without parameter updates. Contribution/Results: The method adapts to distribution shifts online with theoretical guarantees of asymptotic conditional coverage. Extensive multi-task experiments demonstrate significant improvements in both calibration (i.e., validity) and informativeness (i.e., interval width) under small-sample and distribution-shift scenarios, while maintaining low computational overhead.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Planning, Routing, and Scheduling: Temporal Planning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Conformal prediction offers a powerful framework for building distribution-free prediction intervals for exchangeable data. Existing methods that extend conformal prediction to sequential data rely on fitting a relatively complex model to capture temporal dependencies. However, these methods can fail if the sample size is small and often require expensive retraining when the underlying data distribution changes. To overcome these limitations, we propose Reservoir Conformal Prediction (ResCP), a novel training-free conformal prediction method for time series. Our approach leverages the efficiency and representation learning capabilities of reservoir computing to dynamically reweight conformity scores. In particular, we compute similarity scores among reservoir states and use them to adaptively reweight the observed residuals at each step. With this approach, ResCP enables us to account for local temporal dynamics when modeling the error distribution without compromising computational scalability. We prove that, under reasonable assumptions, ResCP achieves asymptotic conditional coverage, and we empirically demonstrate its effectiveness across diverse forecasting tasks.

Problem

• Develop training-free conformal prediction for time series forecasting
• Overcome limitations of complex models with small sample sizes
• Achieve conditional coverage while maintaining computational scalability

Innovation

• Training-free conformal prediction for time series
• Dynamic reweighting using reservoir computing states
• Achieves conditional coverage with computational scalability