Research

2024

1. NeurIPS

   

   Conformal Classification with Equalized Coverage for Adaptively Selected Groups

   Yanfei Zhou, and Matteo Sesia

   In Advances in Neural Information Processing Systems (NeurIPS), 2024

   📄 Paper Summary Code Poster

   This paper introduces a conformal inference method to evaluate uncertainty in classification by generating prediction sets with valid coverage conditional on adaptively chosen features. These features are carefully selected to reflect potential model limitations or biases. This can be useful to find a practical compromise between efficiency—by providing informative predictions—and algorithmic fairness—by ensuring equalized coverage for the most sensitive groups. We demonstrate the validity and effectiveness of this method on simulated and real data sets.

2. ICML

   

   Conformalized Adaptive Forecasting of Heterogeneous Trajectories

   Yanfei Zhou, Lars Lindemann, and Matteo Sesia

   In International Conference on Machine Learning (ICML), 2024

   📄 Paper Summary Code Poster

   We present a new conformal method for generating simultaneous forecasting bands guaranteed to cover the entire path of a new random trajectory with sufficiently high probability. Prompted by the need for dependable uncertainty estimates in motion planning applications where the behavior of diverse objects may be more or less unpredictable, we blend different techniques from online conformal prediction of single and multiple time series and ideas for addressing heteroscedasticity in regression. This solution is both principled, providing precise finite-sample guarantees, and effective, often leading to more informative predictions than prior methods.

2023

1. ICML

   

   Conformal inference is (almost) free for neural networks trained with early stopping

   Ziyi Liang , Yanfei Zhou, and Matteo Sesia

   In International Conference on Machine Learning (ICML), 2023

   📄 Paper Summary Code Poster

   We propose conformalized early stopping, a method that combines early stopping with conformal calibration while efficiently reusing a single hold-out set. This yields models that are accurate and provide reliable uncertainty quantification without multiple data splits. We develop practical implementations for outlier detection, multiclass classification, and regression.

2022

1. NeurIPS

   

   Training uncertainty-aware classifiers with conformalized deep learning

   Bat-Sheva Einbinder, Yaniv Romano, Matteo Sesia , and Yanfei Zhou

   In Advances in Neural Information Processing Systems (NeurIPS), 2022

   📄 Paper Summary Code Poster

   We develop a training strategy for deep multiclass classification that improves model calibration, leading to more informative, reliable uncertainty estimates. Its core is a novel loss function that mitigates overfitting and overconfidence often observed with cross-entropy training. We show the method’s reliability and efficiency on synthetic and real datasets such as CIFAR-10.