publications

2024

1. AISTATS 2024

   Faster Recalibration of an Online Predictor via Approachability

   Okoroafor, Princewill, Sun, Wen, and Kleinberg, Robert

   International Conference on Artificial Intelligence and Statistics (AISTATS 2024) 2024

   Abs PDF

   Predictive models in ML need to be trustworthy and reliable, which often at the very least means outputting calibrated probabilities. This can be particularly difficult to guarantee in the online prediction setting when the outcome sequence can be generated adversarially. In this paper we introduce a technique using Blackwell’s approachability theorem for taking an online predictive model which might not be calibrated and transforming its predictions to calibrated predictions without much increase to the loss of the original model. Our proposed algorithm achieves calibration at a faster rate than existing techniques \citepkuleshov and is the first algorithm to offer a flexible tradeoff between calibration error and accuracy in the online setting. We demonstrate this by characterizing the space of jointly achievable calibration and regret using our technique.

2023

1. Preprint

   Omniprediction for Regression and the Approximate Rank of Convex Functions

   Preprint, 2023

   Abs PDF

   Consider the supervised learning setting where the goal is to learn to predict labels given points from a distribution. An \textitomnipredictor for a class of loss functions and a class of hypotheses is a predictor whose predictions incur less expected loss than the best hypothesis in for every loss in . Since the work of [GKR+21] that introduced the notion, there has been a large body of work in the setting of binary labels where , but much less is known about the regression setting where can be continuous. Our main conceptual contribution is the notion of \textitsufficient statistics for loss minimization over a family of loss functions: these are a set of statistics about a distribution such that knowing them allows one to take actions that minimize the expected loss for any loss in the family. The notion of sufficient statistics relates directly to the approximate rank of the family of loss functions. Our key technical contribution is a bound of on the -approximate rank of convex, Lipschitz functions on the interval , which we show is tight up to a factor of . This yields improved runtimes for learning omnipredictors for the class of all convex, Lipschitz loss functions under weak learnability assumptions about the class . We also give efficient omnipredictors when the loss families have low-degree polynomial approximations, or arise from generalized linear models (GLMs). This translation from sufficient statistics to faster omnipredictors is made possible by lifting the technique of loss outcome indistinguishability introduced by [GKH+23] for Boolean labels to the regression setting.

2. AISTATS 2024

   On the Vulnerability of Fairness Constrained Learning to Malicious Noise

   Blum, Avrim,  Okoroafor, Princewill, Saha, Aadirupa, and Stangl, Kevin

   Neurips Workshop on Algorithmic Fairness, International Conference on Artificial Intelligence and Statistics (AISTATS 2024) 2023

   Abs PDF

   We consider the vulnerability of fairness-constrained learning to small amounts of malicious noise in the training data. Konstantinov and Lampert (2021) initiated the study of this question and presented negative results showing there exist data distributions where for several fairness constraints, any proper learner will exhibit high vulnerability when group sizes are imbalanced. Here, we present a more optimistic view, showing that if we allow randomized classifiers, then the landscape is much more nuanced. For example, for Demographic Parity we show we can incur only a Θ(α) loss in accuracy, where α is the malicious noise rate, matching the best possible even without fairness constraints. For Equal Opportunity, we show we can incur an O(α‾‾√) loss, and give a matching Ω(α‾‾√)lower bound. In contrast, Konstantinov and Lampert (2021) showed for proper learners the loss in accuracy for both notions is Ω(1). The key technical novelty of our work is how randomization can bypass simple "tricks" an adversary can use to amplify his power. We also consider additional fairness notions including Equalized Odds and Calibration. For these fairness notions, the excess accuracy clusters into three natural regimes O(α),O(α‾‾√) and O(1). These results provide a more fine-grained view of the sensitivity of fairness-constrained learning to adversarial noise in training data.

3. SODA 2023

   Non-Stochastic CDF Estimation Using Threshold Queries

   Okoroafor, Princewill, Gupta, Vaishnavi, Kleinberg, Robert, and Goh, Eleanor

   Symposium on Discrete Algorithms, 2023

   Abs PDF

   Estimating a univariate distribution is one of the most basic and fundamental tasks in non-parametric statistics. In this paper, we tackle the distribution estimation problem in a setting with two challenging features. First, the algorithm does not directly observe the data; instead, it only asks a limited number of threshold queries about each sample. Second, the data are not assumed to be independent and identically distributed; instead, we allow for an arbitrary process generating the samples, including an adaptive adversary. These considerations are relevant, for example, when modeling a seller experimenting with posted prices to estimate the distribution of consumers’ willingness to pay for a product: offering a price and observing a consumer’s purchase decision is equivalent to asking a single threshold query about their value, and the distribution of consumers’ values may be non-stationary over time, as early adopters may differ markedly from late adopters. Our main result quantifies, to within a constant factor, the sample complexity of estimating the empirical CDF of a sequence of elements of [n], up to \eps additive error, using one threshold query per sample. The complexity depends only logarithmically on n, and our result can be interpreted as extending the existing logarithmic-complexity results for noisy binary search to the more challenging setting where noise is non-stochastic. Along the way to designing our algorithm, we consider a more general model in which the algorithm is allowed to make a limited number of simultaneous threshold queries on each sample. We solve this problem using Blackwell’s Approachability Theorem and the exponential weights method. As a side result of independent interest, we characterize the minimum number of simultaneous threshold queries required by deterministic CDF estimation algorithms.