Abstract:We are interested in testing properties of distributions with systematically mislabeled samples. Our goal is to make decisions about unknown probability distributions, using a sample that has been collected by a confused collector, such as a machine-learning classifier that has not learned to distinguish all elements of the domain. The confused collector holds an unknown clustering of the domain and an input distribution $\mu$, and provides two oracles: a sample oracle which produces a sample from $\mu$ that has been labeled according to the clustering; and a label-query oracle which returns the label of a query point $x$ according to the clustering. Our first set of results shows that identity, uniformity, and equivalence of distributions can be tested efficiently, under the earth-mover distance, with remarkably weak conditions on the confused collector, even when the unknown clustering is adversarial. This requires defining a variant of the distribution testing task (inspired by the recent testable learning framework of Rubinfeld & Vasilyan), where the algorithm should test a joint property of the distribution and its clustering. As an example, we get efficient testers when the distribution tester is allowed to reject if it detects that the confused collector clustering is "far" from being a decision tree. The second set of results shows that we can sometimes do significantly better when the clustering is random instead of adversarial. For certain one-dimensional random clusterings, we show that uniformity can be tested under the TV distance using $\widetilde O\left(\frac{\sqrt n}{\rho^{3/2} \epsilon^2}\right)$ samples and zero queries, where $\rho \in (0,1]$ controls the "resolution" of the clustering. We improve this to $O\left(\frac{\sqrt n}{\rho \epsilon^2}\right)$ when queries are allowed.

Learning Intersections of Halfspaces with Distribution Shift: Improved Algorithms and SQ Lower Bounds

Testable Learning with Distribution Shift

Tolerant Algorithms for Learning with Arbitrary Covariate Shift

Efficient Discrepancy Testing for Learning with Distribution Shift

Tester-Learners for Halfspaces: Universal Algorithms

Improved Hardness Results for Learning Intersections of Halfspaces

Efficient Testable Learning of General Halfspaces with Adversarial Label Noise

Feature Shift Detection: Localizing Which Features Have Shifted via Conditional Distribution Tests

Revisiting Knowledge Distillation under Distribution Shift

Non-Convex SGD Learns Halfspaces with Adversarial Label Noise

Distributionally robust halfspace depth

Tackling Combinatorial Distribution Shift: A Matrix Completion Perspective

Stable Adversarial Learning under Distributional Shifts

Distribution Testing with a Confused Collector

Learning a Single Neuron Robustly to Distributional Shifts and Adversarial Label Noise

Efficiently Learning Adversarially Robust Halfspaces with Noise

Beyond Discrepancy: A Closer Look at the Theory of Distribution Shift

SQ Lower Bounds for Learning Mixtures of Linear Classifiers

Reliable Learning of Halfspaces under Gaussian Marginals

Distribution-free uncertainty quantification for classification under label shift

Testably Learning Polynomial Threshold Functions