SPLICE: Latent Diffusion over JEPA Embeddings for Conformal Time-Series Inpainting

2026-04-30

AI Summary

This study addresses the lack of reliability guarantees in existing time series imputation methods under limited sample sizes, which hinders their applicability in high-stakes domains such as power systems. To overcome this limitation, the authors propose SPLICE, a novel framework that integrates distribution-free online adaptive conformal inference (ACI) with a diffusion-based imputation model built upon a JEPA self-supervised encoder and a conditional latent-space bridging mechanism—augmented with a flow-matching variant—to achieve both high-efficiency imputation and rigorous coverage guarantees, while enabling cross-domain transfer. Evaluated on 13 load forecasting datasets, SPLICE achieves the lowest average MSE (0.056) and best CRPS (0.161), with empirical coverage rates consistently between 93% and 95%, significantly outperforming current state-of-the-art approaches.

Abstract

Generative models for time-series imputation achieve strong reconstruction accuracy, yet provide no finite-sample reliability guarantees, a critical limitation in power systems where imputed values inform dispatch and planning. We introduce SPLICE (Self-supervised Predictive Latent Inpainting with Conformal Envelopes), a modular framework coupling latent generative imputation with distribution-free, online-adaptive prediction intervals. A JEPA encoder maps daily load segments into a 64-dimensional latent space; a conditional latent bridge with four sampling modes generates candidate gap trajectories; an hourly-conditioned decoder maps back to signal space; and Adaptive Conformal Inference (ACI) wraps the output with coverage-guaranteed prediction bands. The flow-matching variant achieves comparable quality to DDIM in 5--10 ODE steps (5-10x speedup). On thirteen load datasets (nine proprietary, three UCI Electricity, ETTh1), SPLICE achieves the lowest mean Load-only MSE (0.056), winning 9/12 non-degenerate datasets at 91-day gaps and 18/32 across all gap lengths vs. five established baselines, and produces the best CRPS (0.161, -18.3% vs. the strongest competitor). ACI delivers 93--95% empirical coverage, correcting under-coverage failures of up to 7.5 pp observed with static conformal prediction. A pooled JEPA encoder trained on nine feeds transfers to four unseen domains, matching or exceeding per-dataset oracles with only a quick bridge fine-tuning.

Problem

• time-series imputation
• reliability guarantees
• conformal prediction
• power systems
• finite-sample coverage

Innovation

• Conformal Inference
• Latent Diffusion
• JEPA Embeddings
• Time-Series Inpainting
• Adaptive Prediction Intervals