Stable Conformal Prediction Sets

When one observes a sequence of variables $(x_1, y_1), \ldots, (x_n, y_n)$, Conformal Prediction (CP) is a methodology that allows to estimate a confidence set for $y_{n+1}$ given $x_{n+1}$ by merely assuming that the distribution of the data is exchangeable. CP sets have guaranteed coverage for any finite population size $n$. While appealing, the computation of such a set turns out to be infeasible in general, e.g. when the unknown variable $y_{n+1}$ is continuous. The bottleneck is that it is based on a procedure that readjusts a prediction model on data where we replace the unknown target by all its possible values in order to select the most probable one. This requires computing an infinite number of models, which often makes it intractable. In this paper, we combine CP techniques with classical algorithmic stability bounds to derive a prediction set computable with a single model fit. We demonstrate that our proposed confidence set does not lose any coverage guarantees while avoiding the need for data splitting as currently done in the literature. We provide some numerical experiments to illustrate the tightness of our estimation when the sample size is sufficiently large, on both synthetic and real datasets.

[1]  E. Candès,et al.  Testing for outliers with conformal p-values , 2021, The Annals of Statistics.

[2]  Adrian Weller,et al.  Approximating Full Conformal Prediction at Scale via Influence Functions , 2022, AAAI.

[3]  I. Takeuchi,et al.  Root-finding approaches for computing conformal prediction set , 2021, Machine Learning.

[4]  Bruce A. Cox,et al.  Conformal uncertainty sets for robust optimization , 2021, COPA.

[5]  Tianbao Yang,et al.  Stability and Generalization of Stochastic Gradient Methods for Minimax Problems , 2021, ICML.

[6]  Nikita Zhivotovskiy,et al.  Stability and Deviation Optimal Risk Bounds with Convergence Rate O(1/n) , 2021, NeurIPS.

[7]  Regina Barzilay,et al.  Few-shot Conformal Prediction with Auxiliary Tasks , 2021, ICML.

[8]  Martin Jaggi,et al.  Exact Optimization of Conformal Predictors via Incremental and Decremental Learning , 2021, ICML.

[9]  Yao Xie,et al.  Conformal prediction interval for dynamic time-series , 2020, ICML.

[10]  Victor Chernozhukov,et al.  An Exact and Robust Conformal Inference Method for Counterfactual and Synthetic Controls , 2017, Journal of the American Statistical Association.

[11]  B. Efron Resampling Plans and the Estimation of Prediction Error , 2021 .

[12]  Matthew J. Holland Making learning more transparent using conformalized performance prediction , 2020, ArXiv.

[13]  Mihaela van der Schaar,et al.  Discriminative Jackknife: Quantifying Uncertainty in Deep Learning via Higher-Order Influence Functions , 2020, ICML.

[14]  Raef Bassily,et al.  Stability of Stochastic Gradient Descent on Nonsmooth Convex Losses , 2020, NeurIPS.

[15]  Ryan Martin,et al.  Valid distribution-free inferential models for prediction , 2020 .

[16]  I. Takeuchi,et al.  Computing Full Conformal Prediction Set with Approximate Homotopy , 2019, NeurIPS.

[17]  Ryan J. Tibshirani,et al.  Predictive inference with the jackknife+ , 2019, The Annals of Statistics.

[18]  Ichiro Takeuchi,et al.  Safe Grid Search with Optimal Complexity , 2018, ICML.

[19]  Jing Lei,et al.  Fast Exact Conformalization of Lasso using Piecewise Linear Homotopy , 2017, 1708.00427.

[20]  Victor Chernozhukov,et al.  Exact and Robust Conformal Inference Methods for Predictive Machine Learning With Dependent Data , 2018, COLT.

[21]  Alessandro Rinaldo,et al.  Distribution-Free Predictive Inference for Regression , 2016, Journal of the American Statistical Association.

[22]  Henrik Boström,et al.  On the Calibration of Aggregated Conformal Predictors , 2017, COPA.

[23]  Yoram Singer,et al.  Train faster, generalize better: Stability of stochastic gradient descent , 2015, ICML.

[24]  Lars Carlsson,et al.  Aggregated Conformal Prediction , 2014, AIAI Workshops.

[25]  Shai Ben-David,et al.  Understanding Machine Learning: From Theory to Algorithms , 2014 .

[26]  Vladimir Vovk,et al.  Conformal Prediction for Reliable Machine Learning: Theory, Adaptations and Applications , 2014 .

[27]  Göran Falkman,et al.  Inductive conformal anomaly detection for sequential detection of anomalous sub-trajectories , 2013, Annals of Mathematics and Artificial Intelligence.

[28]  Vladimir Vovk,et al.  Cross-conformal predictors , 2012, Annals of Mathematics and Artificial Intelligence.

[29]  Julien Mairal,et al.  Optimization with Sparsity-Inducing Penalties , 2011, Found. Trends Mach. Learn..

[30]  F. Bach,et al.  Optimization with Sparsity-Inducing Penalties (Foundations and Trends(R) in Machine Learning) , 2011 .

[31]  Holger Kantz,et al.  The concept of exchangeability in ensemble forecasting , 2011 .

[32]  Sylvain Arlot,et al.  A survey of cross-validation procedures for model selection , 2009, 0907.4728.

[33]  Mohamed Hebiri,et al.  Sparse conformal predictors , 2009, Stat. Comput..

[34]  Harry Wechsler,et al.  Query by Transduction , 2008, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[35]  Vladimir Vovk,et al.  A tutorial on conformal prediction , 2007, J. Mach. Learn. Res..

[36]  Wen-Liang Hung,et al.  LINEX Loss Functions with Applications to Determining the Optimum Process Parameters , 2007 .

[37]  W. Gasarch,et al.  The Book Review Column 1 Coverage Untyped Systems Simple Types Recursive Types Higher-order Systems General Impression 3 Organization, and Contents of the Book , 2022 .

[38]  R. Tibshirani,et al.  Least angle regression , 2004, math/0406456.

[39]  Harris Papadopoulos,et al.  Inductive Confidence Machines for Regression , 2002, ECML.

[40]  André Elisseeff,et al.  Stability and Generalization , 2002, J. Mach. Learn. Res..

[41]  Vladimir Vovk,et al.  Ridge Regression Confidence Machine , 2001, International Conference on Machine Learning.

[42]  Arthur E. Hoerl,et al.  Ridge Regression: Biased Estimation for Nonorthogonal Problems , 2000, Technometrics.

[43]  J. Hiriart-Urruty,et al.  Convex analysis and minimization algorithms , 1993 .

[44]  John E. Angus,et al.  Regression Estimators: A Comparative Study , 1990 .