Optimal errors and phase transitions in high-dimensional generalized linear models

被引:174
作者
Barbier, Jean [1 ,2 ]
Krzakala, Florent [2 ]
Macris, Nicolas [3 ]
Miolane, Leo [4 ]
Zdeborova, Lenka [5 ,6 ]
机构
[1] Abdus Salaam Int Ctr Theoret Phys, Quantitat Life Sci, I-34151 Trieste, Italy
[2] Univ Paris Diderot, Univ Paris Sci & Lettres, Ecole Normale Super, Lab Phys,CNRS,Sorbonne Univ,Sorbonne Paris Cite, F-75005 Paris, France
[3] Ecole Polytech Fed Lausanne, Sch Comp & Commun Sci, Commun Theory Lab, CH-1015 Lausanne, Switzerland
[4] Univ Paris Sci & Lettres, Dept Informat, Ecole Normale Super, CNRS,Inria, F-75005 Paris, France
[5] Univ Paris Saclay, CNRS, Inst Phys Theor, F-91191 Gif Sur Yvette, France
[6] Univ Paris Saclay, Commissariat Energie Atom, F-91191 Gif Sur Yvette, France
基金
瑞士国家科学基金会; 欧洲研究理事会;
关键词
high-dimensional inference; generalized linear model; Bayesian inference; perceptron; approximate message-passing algorithm; MESSAGE-PASSING ALGORITHMS; STATISTICAL-MECHANICS; STORAGE CAPACITY; SIGNAL RECOVERY; INFORMATION; RETRIEVAL; EQUATIONS; CDMA;
D O I
10.1073/pnas.1802705116
中图分类号
O [数理科学和化学]; P [天文学、地球科学]; Q [生物科学]; N [自然科学总论];
学科分类号
07 ; 0710 ; 09 ;
摘要
Generalized linear models (GLMs) are used in high-dimensional machine learning, statistics, communications, and signal processing. In this paper we analyze GLMs when the data matrix is random, as relevant in problems such as compressed sensing, error-correcting codes, or benchmark models in neural networks. We evaluate the mutual information (or "free entropy") from which we deduce the Bayes-optimal estimation and generalization errors. Our analysis applies to the high-dimensional limit where both the number of samples and the dimension are large and their ratio is fixed. Nonrigorous predictions for the optimal errors existed for special cases of GLMs, e.g., for the perceptron, in the field of statistical physics based on the so-called replica method. Our present paper rigorously establishes those decades-old conjectures and brings forward their algorithmic interpretation in terms of performance of the generalized approximate message-passing algorithm. Furthermore, we tightly characterize, for many learning problems, regions of parameters for which this algorithm achieves the optimal performance and locate the associated sharp phase transitions separating learnable and nonlearnable regions. We believe that this random version of GLMs can serve as a challenging benchmark for multipurpose algorithms.
引用
收藏
页码:5451 / 5460
页数:10
相关论文
共 73 条
[71]   Renyi Information Dimension: Fundamental Limits of Almost Lossless Analog Compression [J].
Wu, Yihong ;
Verdu, Sergio .
IEEE TRANSACTIONS ON INFORMATION THEORY, 2010, 57 (08) :3721-3748
[72]   Bayesian signal reconstruction for 1-bit compressed sensing [J].
Xu, Yingying ;
Kabashima, Yoshiyuki ;
Zdeborova, Lenka .
JOURNAL OF STATISTICAL MECHANICS-THEORY AND EXPERIMENT, 2014,
[73]   Statistical physics of inference: thresholds and algorithms [J].
Zdeborova, Lenka ;
Krzakala, Florent .
ADVANCES IN PHYSICS, 2016, 65 (05) :453-552