Publications

IdePHICS is a new lab in EPFL, having just been created in September 2020. For full list of publication of Prof. Florent Krzakala, please visit florentkrzakala.com

2024

On the atypical solutions of the symmetric binary perceptron

D. Barbier; A. El Alaoui; F. Krzakala; L. Zdeborova 

Journal Of Physics A-Mathematical And Theoretical

2024-05-10

Vol. 57 , num. 19, p. 195202.

DOI : 10.1088/1751-8121/ad3a4a

Gaussian universality of perceptrons with random labels

F. Gerace; F. Krzakala; B. Loureiro; L. Stephan; L. Zdeborova 

Physical Review E

2024-03-08

Vol. 109 , num. 3, p. 034305.

DOI : 10.1103/PhysRevE.109.034305

Statistical mechanics of the maximum-average submatrix problem

V. Erba; F. Krzakala; R. Perez Ortiz; L. Zdeborova 

Journal Of Statistical Mechanics-Theory And Experiment

2024-01-01

Vol. 2024 , num. 1, p. 013403.

DOI : 10.1088/1742-5468/ad1391

2023

Phase diagram of stochastic gradient descent in high-dimensional two-layer neural networks

R. Veiga; L. Stephan; B. Loureiro; F. Krzakala; L. Zdeborova 

Journal Of Statistical Mechanics-Theory And Experiment

2023-11-01

Vol. 2023 , num. 11, p. 114008.

DOI : 10.1088/1742-5468/ad01b1

Multi-layer state evolution under random convolutional design

M. Daniels; C. Gerbelot; F. Krzakala; L. Zdeborova 

Journal Of Statistical Mechanics-Theory And Experiment

2023-11-01

Vol. 2023 , num. 11, p. 114002.

DOI : 10.1088/1742-5468/ad0220

Theoretical characterization of uncertainty in high-dimensional linear classification

L. Clarte; B. Loureiro; F. Krzakala; L. Zdeborova 

Machine Learning-Science And Technology

2023-06-01

Vol. 4 , num. 2, p. 025029.

DOI : 10.1088/2632-2153/acd749

Bayesian reconstruction of memories stored in neural networks from their connectivity

S. Goldt; F. Krzakala; L. Zdeborová; N. Brunel 

PLoS Comput Biol

2023-01-30

Vol. 19 , p. e1010813.

DOI : 10.1371/journal.pcbi.1010813

Fluctuations, bias, variance and ensemble of learners: exact asymptotics for convex losses in high-dimension

B. Loureiro; C. Gerbelot; M. Refinetti; G. Sicuro; F. Krzakala 

Journal Of Statistical Mechanics-Theory And Experiment

2023-11-01

Vol. 2023 , num. 11, p. 114001.

DOI : 10.1088/1742-5468/ad0221

Artificial Neural Network Training on an Optical Processor via Direct Feedback Alignment

K. Müller; J. Launay; I. Poli; M. Filipovich; A. Capelli et al. 

2023 Conference on Lasers and Electro-Optics Europe & European Quantum Electronics Conference (CLEO/Europe-EQEC)

2023

2023 Conference on Lasers and Electro-Optics Europe & European Quantum Electronics Conference (CLEO/Europe-EQEC), Munich, Germany, June 26-30, 2023 .

p. 1-1

DOI : 10.1109/CLEO/Europe-EQEC57999.2023.10231380

Error scaling laws for kernel classification under source and capacity conditions

H. C. Cui; B. Loureiro; F. Krzakala; L. Zdeborová 

IOPscience

2023

num. Mach. Learn.: Sci. Technol. 4 035033.

DOI : 10.1088/2632-2153/acf041

Bayes-optimal Learning of Deep Random Networks of Extensive-width

H. C. Cui; F. Krzakala; L. Zdeborová 

Proceedings of the 40th International Conference on Machine Learning, Honolulu, Hawaii, USA. PMLR 202, 2023

2023

International Conference on Machine Learning, Honolulu, Hawaii, USA, July 23-29, 2023.

Expectation consistency for calibration of neural networks

L. A. Clarte; B. Loureiro; F. Krzakala; L. Zdeborová 

Uncertainty in Artificial Intelligence

2023

Proceedings of the 39th Conference on Uncertainty in Artificial Intelligence (UAI 2023), PMLR 216:443–453., Pittsburgh, PA, USA, August 31-4, 2023.

Compressed sensing with ℓ 0-norm: statistical physics analysis & algorithms for signal recovery

D. Barbier; C. Lucibello; L. Saglietti; F. Krzakala; L. Zdeborova 

2023 Ieee Information Theory Workshop, Itw

2023-01-01

IEEE Information Theory Workshop (ITW), Saint-Malo, FRANCE, Apr 23-28, 2023.

p. 323-328

DOI : 10.1109/ITW55543.2023.10161684

On double-descent in uncertainty quantification in overparametrized models

L. A. Clarte; B. Loureiro; F. Krzakala; L. Zdeborová 

Proceedings of the 26th International Conference on Artificial Intelligence and Statistics (AISTATS)

2023

26th International Conference on Artificial Intelligence and Statistics (AISTATS) , Palau de Congressos, Valencia, Spain, April 25-27, 2023.

p. 7089-7125

Tree-AMP: Compositional Inference with Tree Approximate Message Passing

A. Baker; B. Aubin; F. Krzakala; L. Zdeborová 

Journal of Machine Learning Research

2023

Vol. 24 , num. 57, p. 1–89.

2022

Asymptotic Errors for Teacher-Student Convex Generalized Linear Models (Or: How to Prove Kabashima’s Replica Formula)

C. Gerbelot; A. Abbara; F. Krzakala 

IEEE Transactions on Information Theory

2022-11-17

Vol. 69 , num. 3, p. 1824-1852.

DOI : 10.1109/TIT.2022.3222913

Generalization error rates in kernel regression: the crossover from the noiseless to noisy regime*

H. Cui; B. Loureiro; F. Krzakala; L. Zdeborova 

Journal Of Statistical Mechanics-Theory And Experiment

2022-11-01

Vol. 2022 , num. 11, p. 114004.

DOI : 10.1088/1742-5468/ac9829

Perturbative construction of mean-field equations in extensive-rank matrix factorization and denoising

A. Maillard; F. Krzakala; M. Mézard; L. Zdeborová 

Journal of Statistical Mechanics: Theory and Experiment

2022-08-10

Vol. 2022 , num. 8, p. 083301.

DOI : 10.1088/1742-5468/ac7e4c

Multi-layer State Evolution Under Random Convolutional Design

M. Daniels; C. Gerbelot; F. Krzakala; L. Zdeborová 

NeurIPS Proceedings

2022

Subspace clustering in high-dimensions: Phase transitions & Statistical-to-Computational gap

L. Pesce; B. Loureiro; F. Krzakala; L. Zdeborová 

2022

36th Conference on Neural Information Processing Systems (NeurIPS 2022), New Orleans , November 28-December 9, 2022.

Phase diagram of Stochastic Gradient Descent in high-dimensional two-layer neural networks

R. Veiga; L. Stephan; B. Loureiro; F. Krzakala; L. Zdeborová 

2022

36th Conference on Neural Information Processing Systems (NeurIPS 2022), New Orleans, November 28-December 9, 2022.

Optimal denoising of rotationally invariant rectangular matrices

E. Troiani; V. Erba; F. Krzakala; A. Maillard; L. Zdeborová 

Proceedings of Mathematical and Scientific Machine Learning

2022

Vol. 190 , p. 97-112.

Fluctuations, Bias, Variance & Ensemble of Learners: Exact Asymptotics for Convex Losses in High-Dimension

B. Loureiro; C. Gerbelot; M. Refinetti; G. Sicuro; F. Krzakala 

International Conference On Machine Learning, Vol 162

2022-01-01

38th International Conference on Machine Learning (ICML), Baltimore, MD, Jul 17-23, 2022.

Adversarial Robustness by Design Through Analog Computing And Synthetic Gradients

A. Cappelli; R. Ohana; J. Launay; L. Meunier; I. Poli et al. 

ICASSP 2022 – 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

2022-04-27

ICASSP 2022 – 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Singapore, May 23-27, 2022.

p. 3493-3497

DOI : 10.1109/ICASSP43922.2022.9746671

2021

The Gaussian equivalence of generative models for learning with shallow neural networks

S. Goldt; B. Loureiro; G. Reeves; F. Krzakala; M. Mezard et al. 

Proceedings of the 2nd Mathematical and Scientific Machine Learning Conference

2021-12-16

2nd Mathematical and Scientific Machine Learning Conference, Online, August 16-19, 2021.

Dynamical mean-field theory for stochastic gradient descent in Gaussian mixture classification*

F. Mignacco; F. Krzakala; P. Urbani; A. L. Zdeborova 

Journal Of Statistical Mechanics-Theory And Experiment

2021-12-01

Vol. 2021 , num. 12, p. 124008.

DOI : 10.1088/1742-5468/ac3a80

Generalisation error in learning with random features and the hidden manifold model*

F. Gerace; B. Loureiro; F. Krzakala; M. Mezard; L. Zdeborova 

Journal Of Statistical Mechanics-Theory And Experiment

2021-12-01

Vol. 2021 , num. 12, p. 124013.

DOI : 10.1088/1742-5468/ac3ae6

Construction of optimal spectral methods in phase retrieval

A. Maillard; F. Krzakala; l. Yue; L. Zdeborová 

Proceedings of the 2nd Mathematical and Scientific Machine Learning Conference

2021-08-16

2nd Conference on Mathematical and Scientific Machine Learning (MSML 2021), Lausanne, Suisse, August 16-19, 2021.

p. 693-720

Epidemic mitigation by statistical inference from contact tracing data

A. Baker; I. Biazzo; A. Braunstein; G. Catania; L. Dall’Asta et al. 

Proceedings Of The National Academy Of Sciences Of The United States Of America

2021-08-10

Vol. 118 , num. 32, p. e2106548118.

DOI : 10.1073/pnas.2106548118

Classifying high-dimensional Gaussian mixtures: Where kernel methods fail and neural networks succeed

M. Refinetti; S. Goldt; F. Krzakala; L. Zdeborová 

Proceedings of the 38th International Conference on Machine Learning

2021-07-21

38th International Conference on Machine Learning (ICML), Virtual, July 18-24, 2021.

p. 8936-8947

Learning Gaussian Mixtures with Generalized Linear Models: Precise Asymptotics in High-dimensions

B. Loureiro; G. Sicuro; C. Gerbelot; A. Pacco; F. Krzakala et al. 

Advances in Neural Information Processing Systems 34 pre-proceedings (NeurIPS 2021)

2021

35th Conference on Neural Information Processing Systems (NeurIPS 2021), Online, December 7-10, 2021.

Generalization Error Rates in Kernel Regression: The Crossover from the Noiseless to Noisy Regime

H. C. Cui; B. Loureiro; F. Krzakala; L. Zdeborová 

Advances in Neural Information Processing Systems 34 pre-proceedings (NeurIPS 2021)

2021

35th Conference on Neural Information Processing Systems (NeurIPS 2021), Online, December 7-10, 2021.

Learning curves of generic features maps for realistic datasets with a teacher-student model

B. Loureiro; C. Gerbelot; H. C. Cui; S. Goldt; F. Krzakala et al. 

Advances in Neural Information Processing Systems 34 pre-proceedings (NeurIPS 2021)

2021

35th Conference on Neural Information Processing Systems (NeurIPS 2021), Online, December 7-10, 2021.

p. 16-58

2020

The Spiked Matrix Model With Generative Priors

B. Aubin; B. Loureiro; A. Maillard; F. Krzakala; L. Zdeborova 

IEEE Transactions on Information Theory

2020-10-27

Vol. 67 , num. 2, p. 1156-1181.

DOI : 10.1109/TIT.2020.3033985

Mutual Information and Optimality of Approximate Message-Passing in Random Linear Estimation

J. Barbier; N. Macris; M. Dia; F. Krzakala 

Ieee Transactions On Information Theory

2020-07-01

Vol. 66 , num. 7, p. 4270-4303.

DOI : 10.1109/TIT.2020.2990880

Generalization error in high-dimensional perceptrons: Approaching Bayes error with convex optimization

B. Aubin; F. Krzakala; L. Yue; L. Zdeborová 

Proceeding of the 2020 Advances in Neural Information Processing Systems

2020

Advances in Neural Information Processing Systems, Dec 6, 2020 – Dec 12, 2020.

p. 12199–12210

Phase retrieval in high dimensions: Statistical and computational phase transitions

A. Maillard; B. Loureiro; F. Krzakala; L. Zdeborová 

Proceeding of the 2020 Advances in Neural Information Processing Systems

2020

Advances in Neural Information Processing Systems, Dec 6, 2020 – Dec 12, 2020.

p. 11071–11082

Dynamical mean-field theory for stochastic gradient descent in Gaussian mixture classification

F. Mignacco; F. Krzakala; P. Urbani; L. Zdeborová 

Proceeding of the 2020 Advances in Neural Information Processing Systems

2020

Advances in Neural Information Processing Systems, Dec 6, 2020 – Dec 12, 2020.

p. 9540–955

Direct Feedback Alignment Scales to Modern Deep Learning Tasks and Architectures

J. Launay; I. Poli; F. Boniface; F. Krzakala 

Proceeding of the 2020 Advances in Neural Information Processing Systems

2020

Advances in Neural Information Processing Systems, Dec 6, 2020 – Dec 12, 2020.

p. 9346–9360

Complex Dynamics in Simple Neural Networks: Understanding Gradient Flow in Phase Retrieval

S. Sarao Mannelli; G. Biroli; C. Cammarota; F. Krzakala; P. Urbani et al. 

Proceeding of the 2020 Advances in Neural Information Processing Systems

2020

Advances in Neural Information Processing Systems, Dec 6, 2020 – Dec 12, 2020.

p. 3265–3274

Reservoir Computing meets Recurrent Kernels and Structured Transforms

J. Dong; R. Ohana; M. Rafayelyan; f. Krzakala 

Proceeding of the 2020 Advances in Neural Information Processing Systems

2020

Advances in Neural Information Processing Systems, Dec 6, 2020 – Dec 12, 2020.

p. 16785–16796

2019

Entropy and mutual information in models of deep neural networks

M. Gabrie; A. Manoel; C. Luneau; J. Barbier; N. Macris et al. 

Journal Of Statistical Mechanics-Theory And Experiment

2019-12-01

Vol. 2019 , num. 12, p. 124014.

DOI : 10.1088/1742-5468/ab3430

The committee machine: computational to statistical gaps in learning a two-layers neural network

B. Aubin; A. Maillard; J. Barbier; F. Krzakala; N. Macris et al. 

Journal Of Statistical Mechanics-Theory And Experiment

2019-12-01

Vol. 2019 , num. 12, p. 124023.

DOI : 10.1088/1742-5468/ab43d2

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

J. Barbier; F. Krzakala; N. Macris; L. Miolane; L. Zdeborova 

Proceedings of the National Academy of Sciences

2019-03-19

Vol. 116 , num. 12, p. 5451-5460.

DOI : 10.1073/pnas.1802705116

2018

Entropy and mutual information in models of deep neural networks

M. Gabrie; A. Manoel; C. Luneau; J. Barbier; N. Macris et al. 

Advances In Neural Information Processing Systems 31 (Nips 2018)

2018-01-01

32nd Conference on Neural Information Processing Systems (NIPS), Montreal, CANADA, Dec 02-08, 2018.

The committee machine: Computational to statistical gaps in learning a two-layers neural network

B. Aubin; A. Maillard; J. Barbier; F. Krzakala; N. Macris et al. 

Advances In Neural Information Processing Systems 31 (Nips 2018)

2018-01-01

32nd Conference on Neural Information Processing Systems (NIPS), Montreal, CANADA, Dec 02-08, 2018.

The Mutual Information in Random Linear Estimation Beyond i.i.d. Matrices

J. Barbier; N. Macris; A. Maillard; F. Krzakala 

2018 Ieee International Symposium On Information Theory (Isit)

2018-01-01

IEEE International Symposium on Information Theory (ISIT), Vail, CO, Jun 17-22, 2018.

p. 1390-1394

DOI : 10.1109/ISIT.2018.8437522

2016

Scampi: a robust approximate message-passing framework for compressive imaging

J. Barbier; E. W. Tramel; F. Krzakala 

International Meeting On High-Dimensional Data-Driven Science (Hd3-2015)

2016

International Meeting on High-Dimensional Data-Driven Science (HD3), Kyoto, JAPAN, DEC 14-17, 2015.

p. 012013

DOI : 10.1088/1742-6596/699/1/012013