Abstract:Analysis of low-degree polynomial algorithms is a powerful, newly-popular method for predicting computational thresholds in hypothesis testing problems. One limitation of current techniques for this analysis is their restriction to Bernoulli and Gaussian distributions. We expand this range of possibilities by performing the low-degree analysis of hypothesis testing for the Morris class of exponential families, giving a unified treatment of Gaussian, Poisson, gamma, binomial, negative binomial, and generalized hyperbolic secant distributions. We then give several algorithmic applications. 1. In models where a random signal is observed through an exponential family, the success or failure of low-degree polynomials is governed by the $z$-score overlap, the inner product of $z$-score vectors with respect to the null distribution of two independent copies of the signal. 2. In the same models, testing with low-degree polynomials exhibits channel monotonicity: the above distributions admit a total ordering by computational cost of hypothesis testing, according to a scalar parameter describing how the variance depends on the mean in an exponential family. 3. In a spiked matrix model with a particular non-Gaussian noise distribution, the low-degree prediction is incorrect unless polynomials with arbitrarily large degree in individual matrix entries are permitted. This shows that polynomials summing over self-avoiding walks and variants thereof, as proposed recently by Ding, Hopkins, and Steurer (2020) for spiked matrix models with heavy-tailed noise, are suboptimal for this model. Thus low-degree polynomials appear to offer a tradeoff between robustness and strong performance fine-tuned to specific models, and may struggle with problems requiring an algorithm to first examine the input and then use some intermediate computation to choose from one of several inference subroutines.

Active Learning Polynomial Threshold Functions

Testably Learning Polynomial Threshold Functions

Super Non-singular Decompositions of Polynomials and their Application to Robustly Learning Low-degree PTFs

Attribute-Efficient PAC Learning of Low-Degree Polynomial Threshold Functions with Nasty Noise

Polynomial Threshold Functions: Structure, Approximation and Pseudorandomness

Low-degree learning and the metric entropy of polynomials

Heuristic algorithms for obtaining Polynomial Threshold Functions with low densities

Differentially Private Release and Learning of Threshold Functions

Approximating the Number of Relevant Variables in a Parity Implies Proper Learning

Statistical Active Learning Algorithms for Noise Tolerance and Differential Privacy

Efficiently Learning One-Hidden-Layer ReLU Networks via Schur Polynomials

Identifying good directions to escape the NTK regime and efficiently learn low-degree plus sparse polynomials

Iterative thresholding for non-linear learning in the strong $\varepsilon$-contamination model

Learning Thresholds with Latent Values and Censored Feedback

Hypothesis testing with low-degree polynomials in the Morris class of exponential families

Differentially Private Iterative Gradient Hard Thresholding for Sparse Learning

Polynomial Bounds for Learning Noisy Optical Physical Unclonable Functions and Connections to Learning With Errors

Learning low-degree quantum objects

Noninteractive Locally Private Learning of Linear Models via Polynomial Approximations

Accurate Low-Degree Polynomial Approximation of Non-polynomial Operators for Fast Private Inference in Homomorphic Encryption

Low coordinate degree algorithms I: Universality of computational thresholds for hypothesis testing