A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification

2021/07/15 by Anastasios N. Angelopoulos, Stephen Bates, Angelopoulos, Anastasios N. +1 · 3 voices · 147 citations
Computer Science · #Anomaly Detection Techniques and Applications #Explainable Artificial Intelligence (XAI) #Adversarial Robustness in Machine Learning

paper · pdf · doi:10.48550/arxiv.2107.07511

Abstract

Black-box machine learning models are now routinely used in high-risk settings, like medical diagnostics, which demand uncertainty quantification to avoid consequential model failures. Conformal prediction is a user-friendly paradigm for creating statistically rigorous uncertainty sets/intervals for the predictions of such models. Critically, the sets are valid in a distribution-free sense: they possess explicit, non-asymptotic guarantees even without distributional assumptions or model assumptions. One can use conformal prediction with any pre-trained model, such as a neural network, to produce sets that are guaranteed to contain the ground truth with a user-specified probability, such as 90%. It is easy-to-understand, easy-to-use, and general, applying naturally to problems arising in the fields of computer vision, natural language processing, deep reinforcement learning, and so on. This hands-on introduction is aimed to provide the reader a working understanding of conformal prediction and related distribution-free uncertainty quantification techniques with one self-contained document. We lead the reader through practical theory for and examples of conformal prediction and describe its extensions to complex machine learning tasks involving structured outputs, distribution shift, time-series, outliers, models that abstain, and more. Throughout, there are many explanatory illustrations, examples, and code samples in Python. With each code sample comes a Jupyter notebook implementing the method on a real-data example; the notebooks can be accessed and easily run using our codebase.

Citations

• Distribution-free, Risk-controlling Prediction Sets

Cited by

• Combining gravitational wave search pipelines to find subthreshold signals in GWTC-5.0
• pacSTL: PAC-Bounded Signal Temporal Logic from Data-Driven Reachability Analysis
• When to Trust the Map: Confidence-Aware LLM Routing for Automotive CVE-to-ATM Mapping
• AEVAL: From Anecdotal to Deterministic Testing for Agentic Skill Workflows
• Signed Evidence Flow: Conflict-Aware and Stability-Calibrated Data Analysis
• Falsifiable Release Gates for Self-Improving Systems: Standing Invariants at Scale
• Heteroskedastic Signals in Budgeted LLM Verification: Structural Heterogeneity Limits Optimization Gains
• Multi-Granularity Conformal Prediction for Reliable Neural-Operator Automotive Aerodynamic Surrogates
• Evaluating Conformal Reliability of Pathway-Level Transcriptomic Signatures Under Cross-Cohort Shift in Sepsis Mortality Prediction
• Spectral Adaptive Conformal Prediction for Structured Non-Exchangeable Data
• Contrastive Conformal Sets
• Coverage Guarantees for Pseudo-Calibrated Conformal Prediction under Distribution Shift
• Drift-Aware Spectral Conformal Prediction for Non-Exchangeable Streaming Data
• A Step Forward Towards Trustworthy Risk-Aware Facial Retrieval (RA-FR)
• When Is Delegated Play Truthful? Within-Range Regret and the Trilemma of Aligned Delegation
• Cloud-Native Evaluation-as-a-Service: A Microservices Architecture for Scalable AI Monitoring with Conformal Guarantees
• A Collectivist, Economic Perspective on AI
• Distribution-Free Process Monitoring with Conformal Prediction
• proxymate: Diagnosis and Adjustment of Proxy Estimates for Reliable Inference
• Conformal predictive intervals in survival analysis: a re-sampling approach
• Vehicle-to-Everything Cooperative Perception for Autonomous Driving
• FinAbstain: Uncertainty-Calibrated Multimodal RAG for Selective Financial Forecasting
• Gradient-Based Latent Decomposition Reveals Mechanisms of Feature Degradation in Weakly Supervised Mammography
• A Diagnostic Gap Framework for Evaluating Reconstruction Fidelity in Weakly Supervised Mammography
• Efficient Online Conformal Selection with Limited Feedback
• Enhancing Decision-Making in Windows PE Malware Classification During Dataset Shifts with Uncertainty Estimation
• Another Fit Bites the Dust: Conformal Prediction as a Calibration Standard for Machine Learning in High-Energy Physics
• The Semantic Illusion: Certified Limits of Embedding-Based Hallucination Detection in RAG Systems
• Weighted Conformal Prediction Provides Adaptive and Valid Mask-Conditional Coverage for General Missing Data Mechanisms
• General OOD Detection via Model-aware and Subspace-aware Variable Priority
• Selective Conformal Risk Control
• LYNX: Learning Dynamic Exits for Confidence-Controlled Reasoning
• Seeing Soil from Space: Towards Robust and Scalable Remote Soil Nutrient Analysis
• A Critical Perspective on Finite Sample Conformal Prediction Theory in Medical Applications
• Symmetric Aggregation of Conformity Scores for Efficient Uncertainty Sets
• Posterior inference via Hill's prediction model
• Reliable Statistical Guarantees for Conformal Predictors with Small Datasets
• Pick-to-Learn for Systems and Control: Data-driven Synthesis with State-of-the-art Safety Guarantees
• Modal Logical Neural Networks
• LEC: Linear Expectation Constraints for Selection-Conditioned Risk Control in Selective Prediction and Routing Systems
• What AI Speaks for Your Community: Polling AI Agents for Public Opinion on Data Center Projects
• Sample-Efficient Expert Query Control in Active Imitation Learning via Conformal Prediction
• Beyond Accuracy: An Empirical Study of Uncertainty Estimation in Imputation
• Conformal Safety Monitoring for Flight Testing: A Case Study in Data-Driven Safety Learning
• Safe and Stable Neural Network Dynamical Systems for Robot Motion Planning
• Enhancing Conformal Prediction via Class Similarity
• Hierarchical biomarker thresholding: a model-agnostic framework for stability
• Conformal Prediction for Compositional Data
• Principled Context Engineering for RAG: Statistical Guarantees via Conformal Prediction
• Cost-Sensitive Conformal Training with Provably Controllable Learning Bounds
• Provably Minimum-Length Conformal Prediction Sets for Ordinal Classification
• Attention-Based Feature Online Conformal Prediction for Time Series
• Controlling False Positives in Image Segmentation via Conformal Prediction
• Conformal Online Learning of Deep Koopman Linear Embeddings
• Calibrated Decomposition of Aleatoric and Epistemic Uncertainty in Deep Features for Inference-Time Adaptation
• Safe Planning in Interactive Environments via Iterative Policy Updates and Adversarially Robust Conformal Prediction
• Semi-Supervised Treatment Effect Estimation with Unlabeled Covariates via Generalized Riesz Regression
• Statistically Assuring Safety of Control Systems using Ensembles of Safety Filters and Conformal Prediction
• CoFineLLM: Conformal Finetuning of LLMs for Language-Instructed Robot Planning
• Uncertainty Quantification for Reduced-Order Surrogate Models Applied to Cloud Microphysics
• Conformal Prediction-Driven Adaptive Sampling for Digital Twins of Water Distribution Networks
• Relational Deep Dive: Error-Aware Queries Over Unstructured Data
• Online Energy Storage Arbitrage under Imperfect Predictions: A Conformal Risk-Aware Approach
• Dynamic Model Selection for Trajectory Prediction via Pairwise Ranking and Meta-Features
• Before the Clinic: Transparent and Operable Design Principles for Healthcare AI
• A systematic evaluation of uncertainty quantification techniques in deep learning: a case study in photoplethysmography signal analysis
• Simultaneous Coverage and Efficiency Guarantee in Online Conformal Prediction
• Towards Trustworthy Embodied Intelligence: A Systems Framework and Graded Trustworthiness Levels
• Structurally Separated Uncertainty in Supervised Latent Variable Models
• LRT-Diffusion: Calibrated Risk-Aware Guidance for Diffusion Policies
• Conformal Prediction in The Loop: A Feedback-Based Uncertainty Model for Trajectory Optimization
• Robust Uncertainty Quantification for Self-Evolving Large Language Models via Continual Domain Pretraining
• Multi-Modal Fact-Verification Framework for Reducing Hallucinations in Large Language Models
• What Does It Take to Build a Performant Selective Classifier?
• Foundation Model Forecasts: Form and Function
• Extending Prediction-Powered Inference through Conformal Prediction
• Adaptive Individual Uncertainty under Out-Of-Distribution Shift with Expert-Routed Conformal Prediction
• Federated Conditional Conformal Prediction via Generative Models
• Km-scale dynamical downscaling through conformalized latent diffusion models
• UNCAP: Uncertainty-Guided Planning Using Natural Language Communication for Cooperative Autonomous Vehicles
• Simultaneous Frequentist Calibration of Confidence Regions for Multiple Functionals in Constrained Inverse Problems
• Uncertainty-Aware, Risk-Adaptive Access Control for Agentic Systems using an LLM-Judged TBAC Model
• Modelling Territorial Dynamics through Statistical Mechanics: An Application to Resident Foreign Population
• metabeta -- A fast neural model for Bayesian mixed-effects regression
• Valid Stopping for LLM Generation via Empirical Dynamic Formal Lift
• INSIGHT: INference-time Sequence Introspection for Generating Help Triggers in Vision-Language-Action Models
• Domain-Shift-Aware Conformal Prediction for Large Language Models
• Uncertainty in Machine Learning
• Relevance-Aware Thresholding in Online Conformal Prediction for Time Series
• Annotation-Efficient Active Test-Time Adaptation with Conformal Prediction
• Safe Planning in Unknown Environments using Conformalized Semantic Maps
• Uncertainty-Guided Expert-AI Collaboration for Efficient Soil Horizon Annotation
• Conformal Prediction for Signal Temporal Logic Inference
• Uncertainty quantification in atomistic simulations of silicon using interatomic potentials
• Taming Variability: Randomized and Bootstrapped Conformal Risk Control for LLMs
• Structure and transport properties of LiTFSI-based deep eutectic electrolytes from machine-learned interatomic potential simulations
• Smoothing-Based Conformal Prediction for Balancing Efficiency and Interpretability
• COMPASS: Robust Feature Conformal Prediction for Medical Segmentation Metrics
• Unsupervised Conformal Inference: Bootstrapping and Alignment to Control LLM Uncertainty
• Conditional predictive inference with L^k-coverage control
• A Systematic Review of Conformal Inference Procedures for Treatment Effect Estimation: Methods and Challenges
• Filtering with Confidence: When Data Augmentation Meets Conformal Prediction
• Document Summarization with Conformal Importance Guarantees
• ConRad: Efficient Conformal Prediction for Radiomics
• Geometric Risk Control for Vision-Language Model OCR
• Analyzing Uncertainty of LLM-as-a-Judge: Interval Evaluations with Conformal Prediction
• Are you sure? Measuring models bias in content moderation through uncertainty
• A Conformal Prediction Framework for Uncertainty Quantification in Physics-Informed Neural Networks
• The Art of Saying "Maybe": A Conformal Lens for Uncertainty Benchmarking in VLMs
• No Need for Learning to Defer? A Training Free Deferral Framework to Multiple Experts through Conformal Prediction
• Selective Risk Certification for LLM Outputs via Information-Lift Statistics: PAC-Bayes, Robustness, and Skeleton Design
• Probabilistic Robustness Analysis in High Dimensional Space: Application to Semantic Segmentation Network
• Feasibility-Guided Fair Adaptive Offline Reinforcement Learning for Medicaid Care Management
• Measuring Uncertainty in Transformer Circuits with Effective Information Consistency
• Statistical Methods in Generative AI
• Performance of Conformal Prediction in Capturing Aleatoric Uncertainty
• Fundamental bounds on efficiency-confidence trade-off for transductive conformal prediction
• Conformal Prediction for Time-series Forecasting with Change Points
• Towards Trustworthy Vital Sign Forecasting: Leveraging Uncertainty for Prediction Intervals
• Energy Landscapes Enable Reliable Abstention in Retrieval-Augmented Large Language Models for Healthcare
• CUTE-MRI: Conformalized Uncertainty-based framework for Time-adaptivE MRI
• Effect of Data Augmentation on Conformal Prediction for Diabetic Retinopathy
• TCUQ: Single-Pass Uncertainty Quantification from Temporal Consistency with Streaming Conformal Calibration for TinyML
• Uncertainty-Aware Prediction of Parkinson's Disease Medication Needs: A Two-Stage Conformal Prediction Approach
• Scalable, Technology-Agnostic Diagnosis and Predictive Maintenance for Point Machine using Deep Learning
• CVCM Track Circuits Pre-emptive Failure Diagnostics for Predictive Maintenance Using Deep Neural Networks
• Membership Inference Attacks with False Discovery Rate Control
• Conformal P-Value in Multiple-Choice Question Answering Tasks with Provable Risk Control
• Conformal Sets in Multiple-Choice Question Answering under Black-Box Settings with Provable Coverage Guarantees
• Conformal Unlearning: A New Paradigm for Unlearning in Conformal Predictors
• Random-Access Ranked Retrieval and Similarity Search
• ConformalSAM: Unlocking the Potential of Foundational Segmentation Models in Semi-Supervised Semantic Segmentation with Conformal Prediction
• Safety Evaluation of Motion Plans Using Trajectory Predictors as Forward Reachable Set Estimators
• Multi-Agent Path Finding Among Dynamic Uncontrollable Agents with Statistical Safety Guarantees
• Approximating Full Conformal Prediction for Neural Network Regression with Gauss-Newton Influence
• Programmable Virtual Humans Toward Human Physiologically-Based Drug Discovery
• Auto: The AGI Compiler
• QUTCC: Quantile Uncertainty Training and Conformal Calibration for Imaging Inverse Problems
• AEGIS: A Backup Reflex for Physical AI
• Heterogeneous Predictor-based Risk-Aware Planning with Conformal Prediction in Dense, Uncertain Environments
• Differentially Private Conformal Prediction via Quantile Binary Search
• LLM-based ambiguity detection in natural language instructions for collaborative surgical robots
• Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Control
• Non-exchangeable Conformal Prediction with Optimal Transport: Tackling Distribution Shifts with Unlabeled Data
• Foundation models for time series forecasting: Application in conformal prediction
• Conformal Prediction for Long-Tailed Classification
• Machine Learning in Acoustics: A Review and Open-Source Repository

Discussions

• Ever run a complicated supervised learning alg., need to quantify uncertainty, but have no idea how to calculate standard errors? Conformal prediction provides a simple, versatile, finite-sample-valid [bsky, 26 points, 1 comments]
• little known fact, you can train a model to have a specific FNR ("conformal risk control") arxiv.org/abs/2107.07511 [bsky, 9 points, 0 comments]
• I looked it up, this is a nice paper on it: arxiv.org/pdf/2107.07511 [bsky, 0 points, 0 comments]

Related

• A comparative study of conformal prediction methods for valid uncertainty quantification in machine learning
• Conformal Prediction: A Data Perspective
• Conformal Prediction: A Data Perspective
• A Critical Perspective on Finite Sample Conformal Prediction Theory in Medical Applications
• Conformal Prediction as Bayesian Quadrature