v266: Proceedings of COPA 2025

2026-06-04 · Source: Proceedings of Machine Learning Research · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, medium

Summary

The Fourteenth Symposium on Conformal and Probabilistic Prediction with Applications (COPA 2025) proceedings, edited by Khuong An Nguyen, Zhiyuan Luo, Harris Papadopoulos, Tuwe Löfström, Lars Carlsson, and Henrik Boström, compiles research presented from 10-12 September 2025 at Royal Holloway, London, UK. This volume, PMLR 266, covers diverse areas including foundational advancements like inductive randomness predictors and calibration set reuse, alongside critical aspects of privacy, robustness, and security, such as local differential privacy and handling noisy labels. It also explores multi-output and multi-label prediction, conformal regression, and predictive distributions, introducing methods like SEMF and stacked conformal prediction. Significant sections address explainability, decision-making, and a wide array of applications, from pulmonary nodule detection and autonomous driving to drug-target interaction prediction and stock selections. The symposium further delves into online testing, change, and anomaly detection, featuring tools like the online-cp Python package and methods for recurrent concept drift.

Key takeaway

For AI Scientists and Research Scientists integrating uncertainty quantification into machine learning models, this symposium's proceedings offer a comprehensive resource. You should explore the advancements in Conformal Prediction for enhancing model reliability and explainability across various domains. Consider applying specific techniques for privacy-preserving predictions, multi-output regression, or online anomaly detection to improve your model's robustness, interpretability, and decision-making capabilities in critical applications.

Key insights

Conformal Prediction provides statistically rigorous uncertainty quantification across diverse machine learning applications and challenges.

Principles

• Conformal Prediction offers provable validity for uncertainty quantification.
• Calibration sets can be reused for multiple predictions under specific conditions.
• Privacy-preserving and robust CP methods are essential for sensitive data.

Method

Source-free conformal prediction can be achieved using pseudo-labels. Dynamic Conformal Prediction optimizes informational efficiency. SHAP analysis explains set-valued conformal predictions.

In practice

• Apply Conformal Prediction for reliable pulmonary nodule detection.
• Utilize CP for uncertainty estimation in drug-target interaction prediction.
• Employ the online-cp Python package for real-time anomaly detection.

Topics

• Conformal Prediction
• Uncertainty Quantification
• Machine Learning Robustness
• Privacy-Preserving AI
• Anomaly Detection
• Multi-Output Prediction
• Explainable AI

Code references

• egonmedhatten/online-cp

Best for: AI Scientist, Research Scientist

Related on AIssential

• [D] Conformal Prediction vs naive thresholding to represent uncertainty — Conformal Prediction offers theoretical guarantees for uncertainty quantification, unlike naive thresholding.
• Distribution-Free Uncertainty Quantification for Continuous AI Agent Evaluation — Conformal prediction offers robust, distribution-free uncertainty quantification for continuous AI agent evaluation.
• Online Conformal Prediction: Enforcing monotonicity via Online Optimization — New online conformal prediction methods ensure nested prediction sets across multiple confidence levels.
• The ninety-percent confidence that means nothing — and what does — Softmax outputs are not true probabilities and misrepresent classifier confidence; conformal prediction offers reliable coverage guarantees.
• Beyond Surface Statistics: Robust Conformal Prediction for LLMs via Internal Representations — Internal LLM representations offer more robust uncertainty signals for conformal prediction, especially under distribution shifts.
• Partial Causal Structure Learning for Valid Selective Conformal Inference under Interventions — Selective conformal prediction under interventions can be robustly maintained by learning partial causal structures.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Proceedings of Machine Learning Research.