Marvels and pitfalls of the Langevin algorithm in noisy high-dimensional inference

Autor: Lenka Zdeborová, Florent Krzakala, Chiara Cammarota, Pierfrancesco Urbani, Giulio Biroli, Stefano Sarao Mannelli

Publikováno v: Cammarota, C, Sarao, S, Biroli, G, Krzakala, F, Zdeborova, L & Urbani, P 2020, ' Marvels and pitfalls of the Langevin algorithm in noisy high-dimensional inference ', Physical Review X, vol. 10, no. 1, 011057, pp. 011057-1-011057-41 . https://doi.org/10.1103/PhysRevX.10.011057
Physical Review X
Physical Review X, American Physical Society, 2020, 10 (011057)
Physical Review X, Vol 10, Iss 1, p 011057 (2020)
Physical Review X, 2020, 10, pp.011057

Gradient-descent-based algorithms and their stochastic versions have widespread applications in machine learning and statistical inference. In this work we perform an analytic study of the performances of one of them, the Langevin algorithm, in the c

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::8cfee70840f568aba77ca390ba26e68e
https://kclpure.kcl.ac.uk/en/publications/84bc9c04-f4b7-4db0-971a-fd29a083f3ca

Thresholds of descending algorithms in inference problems.

Autor: Stefano Sarao Mannelli, Lenka Zdeborová

Publikováno v: Journal of Statistical Mechanics: Theory & Experiment; Mar2020, Vol. 2020 Issue 3, p1-1, 1p

An analytical theory of curriculum learning in teacher–student networks*

Autor: Luca Saglietti, Stefano Sarao Mannelli, Andrew Saxe

Publikováno v: Journal of Statistical Mechanics: Theory and Experiment. 2022(11):114014

In humans and animals, curriculum learning -- presenting data in a curated order - is critical to rapid learning and effective pedagogy. Yet in machine learning, curricula are not widely used and empirically often yield only moderate benefits. This s

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::4e283f52e608de31c6bf331376e28e47