analiza aiqbench
This research introduces Local Sliced Conformal Inference (LSCI), a distribution-free framework that provides tighter, locally adaptive prediction sets for operator models used in spatiotemporal forecasting. By proving finite-sample validity and demonstrating robustness against biased predictions, the method offers improved uncertainty quantification for high-stakes applications like weather prediction and energy demand modeling. The approach addresses a critical need for calibrated confidence intervals in complex function-valued regression tasks where standard conformal baselines may lack adaptivity.
Locally Adaptive Conformal Inference for Operator Models
arXiv:2507.20975v5 Announce Type: replace-cross Abstract: Operator models are regression algorithms between Banach spaces of functions. They have become an increasingly critical tool for spatiotemporal forecasting and physics emulation, especially in high-stakes scenarios where robust, calibrated uncertainty quantification is required. We introduce Local Sliced Conformal Inference (LSCI), a distribution-free framework for generating function-valued, locally adaptive prediction sets for operator models. We prove finite-sample validity and derive a data-dependent upper bound on the coverage gap under local exchangeability. On synthetic Gaussian-process tasks and real applications (air quality monitoring, energy demand forecasting, and weather prediction), LSCI yields tighter sets with stronger adaptivity compared to conformal baselines. We also empirically demonstrate robustness against biased predictions and certain out-of-distribution noise regimes.
Abstract:Operator models are regression algorithms between Banach spaces of functions. They have become an increasingly critical tool for spatiotemporal forecasting and physics emulation, especially in high-stakes scenarios where robust, calibrated uncertainty quantification is required. We introduce Local Sliced Conformal Inference (LSCI), a distribution-free framework for generating function-valued, locally adaptive prediction sets for operator models. We prove finite-sample validity and derive a data-dependent upper bound on the coverage gap under local exchangeability. On synthetic Gaussian-process tasks and real applications (air quality monitoring, energy demand forecasting, and weather prediction), LSCI yields tighter sets with stronger adaptivity compared to conformal baselines. We also empirically demonstrate robustness against biased predictions and certain out-of-distribution noise regimes.
Machine Learning (stat.ML); Machine Learning (cs.LG)
arXiv:2507.20975 [stat.ML]
(or arXiv:2507.20975v5 [stat.ML] for this version)
Text integral sub licența deschisă a publicației · 125 cuvinte
Subiecte conexe
Beyond Point Estimates: Benchmarking Uncertainty Quantification Methods on the AION-1 Astronomical Foundation Model
arXiv · 9 iun. 2026
Natera’s intelligent appointment scheduling with Amazon Bedrock AgentCore
AWS · 26 aug. 2026
How to Fine-Tune Nemotron 3.5 ASR for Your Language, Domain, or Accent
Hugging Face · 4 iun. 2026
Deploying Qwen3.8-2.4T-A95B on Amazon SageMaker HyperPod with vLLM
AWS · 9 sept. 2026