Attention-Based Feature Online Conformal Prediction for Time Series

2025/11/19 by Zhu, Meiyi, Guo, Caili, Feng, Chunyan +1
#FOS: Computer and information sciences #FOS: Electrical engineering #Information Theory (cs.IT) #Machine Learning (cs.LG) #Signal Processing (eess.SP) #electronic engineering #information engineering

paper · doi:10.48550/arxiv.2511.15838

Abstract

Online conformal prediction (OCP) wraps around any pre-trained predictor to produce prediction sets with coverage guarantees that hold irrespective of temporal dependencies or distribution shifts. However, standard OCP faces two key limitations: it operates in the output space using simple nonconformity (NC) scores, and it treats all historical observations uniformly when estimating quantiles. This paper introduces attention-based feature OCP (AFOCP), which addresses both limitations through two key innovations. First, AFOCP operates in the feature space of pre-trained neural networks, leveraging learned representations to construct more compact prediction sets by concentrating on task-relevant information while suppressing nuisance variation. Second, AFOCP incorporates an attention mechanism that adaptively weights historical observations based on their relevance to the current test point, effectively handling non-stationarity and distribution shifts. We provide theoretical guarantees showing that AFOCP maintains long-term coverage while provably achieving smaller prediction intervals than standard OCP under mild regularity conditions. Extensive experiments on synthetic and real-world time series datasets demonstrate that AFOCP consistently reduces the size of prediction intervals by as much as 88% as compared to OCP, while maintaining target coverage levels, validating the benefits of both feature-space calibration and attention-based adaptive weighting.

Citations

• Comparing Uncertainty Measurement and Mitigation Methods for Large Language Models: A Systematic Review
• Conformal Calibration: Ensuring the Reliability of Black-Box AI in Wireless Systems
• Theoretical Foundations of Conformal Prediction
• Localized Adaptive Risk Control
• Calibrating Bayesian Learning via Regularization, Confidence Minimization, and Selective Inference
• Group-Weighted Conformal Prediction
• Cross-Validation Conformal Risk Control
• Forking Uncertainties: Reliable Prediction and Model Predictive Control with Sequence Models via Conformal Risk Control
• Conformal PID Control for Time Series Prediction
• Conformal Prediction for Time Series with Modern Hopfield Networks
• Improved Online Conformal Prediction via Strongly Adaptive Online Learning
• Sequential Predictive Conformal Inference for Time Series
• Safe Planning in Dynamic Environments using Conformal Prediction
• Predictive Inference with Feature Conformal Prediction
• Conformal Inference for Online Prediction with Arbitrary Distribution Shifts
• Language Models (Mostly) Know What They Know
• Conformal prediction beyond exchangeability
• Adaptive Conformal Predictions for Time Series
• Multi-resolution spatio-temporal prediction with application to wind power generation
• A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification
• A Survey of Uncertainty in Deep Neural Networks
• Adaptive Conformal Inference Under Distribution Shift
• Conformal Prediction Under Covariate Shift
• On Calibration of Modern Neural Networks
• Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
• Distribution-Free Predictive Inference For Regression
• Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning

Related

• Relevance-Aware Thresholding in Online Conformal Prediction for Time Series
• Mirror Online Conformal Prediction with Intermittent Feedback
• Improved Online Conformal Prediction via Strongly Adaptive Online Learning
• Feature Fitted Online Conformal Prediction for Deep Time Series Forecasting Model
• Exploring the Noise Robustness of Online Conformal Prediction