Complex Energy Landscapes in Spiked-Tensor and Simple Glassy Models: Ruggedness, Arrangements of Local Minima, and Phase Transitions

Valentina Ros, Gerard Ben Arous, Giulio Biroli, Chiara Cammarota

Research output: Contribution to journalArticle

Abstract

We study rough high-dimensional landscapes in which an increasingly stronger preference for a given configuration emerges. Such energy landscapes arise in glass physics and inference. In particular, we focus on random Gaussian functions and on the spiked-tensor model and generalizations. We thoroughly analyze the statistical properties of the corresponding landscapes and characterize the associated geometrical phase transitions. In order to perform our study, we develop a framework based on the Kac-Rice method that allows us to compute the complexity of the landscape, i.e., the logarithm of the typical number of stationary points and their Hessian. This approach generalizes the one used to compute rigorously the annealed complexity of mean-field glass models. We discuss its advantages with respect to previous frameworks, in particular, the thermodynamical replica method, which is shown to lead to partially incorrect predictions.

Original languageEnglish (US)
Article number011003
JournalPhysical Review X
Volume9
Issue number1
DOIs
StatePublished - Jan 4 2019

Fingerprint

ruggedness
tensors
energy
glass
rice
logarithms
replicas
inference
physics
configurations
predictions

ASJC Scopus subject areas

  • Physics and Astronomy(all)

Cite this

Complex Energy Landscapes in Spiked-Tensor and Simple Glassy Models : Ruggedness, Arrangements of Local Minima, and Phase Transitions. / Ros, Valentina; Ben Arous, Gerard; Biroli, Giulio; Cammarota, Chiara.

In: Physical Review X, Vol. 9, No. 1, 011003, 04.01.2019.

Research output: Contribution to journalArticle

@article{6416b5f5a9fd40a1bc8df137df410ea2,
title = "Complex Energy Landscapes in Spiked-Tensor and Simple Glassy Models: Ruggedness, Arrangements of Local Minima, and Phase Transitions",
abstract = "We study rough high-dimensional landscapes in which an increasingly stronger preference for a given configuration emerges. Such energy landscapes arise in glass physics and inference. In particular, we focus on random Gaussian functions and on the spiked-tensor model and generalizations. We thoroughly analyze the statistical properties of the corresponding landscapes and characterize the associated geometrical phase transitions. In order to perform our study, we develop a framework based on the Kac-Rice method that allows us to compute the complexity of the landscape, i.e., the logarithm of the typical number of stationary points and their Hessian. This approach generalizes the one used to compute rigorously the annealed complexity of mean-field glass models. We discuss its advantages with respect to previous frameworks, in particular, the thermodynamical replica method, which is shown to lead to partially incorrect predictions.",
author = "Valentina Ros and {Ben Arous}, Gerard and Giulio Biroli and Chiara Cammarota",
year = "2019",
month = "1",
day = "4",
doi = "10.1103/PhysRevX.9.011003",
language = "English (US)",
volume = "9",
journal = "Physical Review X",
issn = "2160-3308",
publisher = "American Physical Society",
number = "1",

}

TY - JOUR

T1 - Complex Energy Landscapes in Spiked-Tensor and Simple Glassy Models

T2 - Ruggedness, Arrangements of Local Minima, and Phase Transitions

AU - Ros, Valentina

AU - Ben Arous, Gerard

AU - Biroli, Giulio

AU - Cammarota, Chiara

PY - 2019/1/4

Y1 - 2019/1/4

N2 - We study rough high-dimensional landscapes in which an increasingly stronger preference for a given configuration emerges. Such energy landscapes arise in glass physics and inference. In particular, we focus on random Gaussian functions and on the spiked-tensor model and generalizations. We thoroughly analyze the statistical properties of the corresponding landscapes and characterize the associated geometrical phase transitions. In order to perform our study, we develop a framework based on the Kac-Rice method that allows us to compute the complexity of the landscape, i.e., the logarithm of the typical number of stationary points and their Hessian. This approach generalizes the one used to compute rigorously the annealed complexity of mean-field glass models. We discuss its advantages with respect to previous frameworks, in particular, the thermodynamical replica method, which is shown to lead to partially incorrect predictions.

AB - We study rough high-dimensional landscapes in which an increasingly stronger preference for a given configuration emerges. Such energy landscapes arise in glass physics and inference. In particular, we focus on random Gaussian functions and on the spiked-tensor model and generalizations. We thoroughly analyze the statistical properties of the corresponding landscapes and characterize the associated geometrical phase transitions. In order to perform our study, we develop a framework based on the Kac-Rice method that allows us to compute the complexity of the landscape, i.e., the logarithm of the typical number of stationary points and their Hessian. This approach generalizes the one used to compute rigorously the annealed complexity of mean-field glass models. We discuss its advantages with respect to previous frameworks, in particular, the thermodynamical replica method, which is shown to lead to partially incorrect predictions.

UR - http://www.scopus.com/inward/record.url?scp=85062007999&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85062007999&partnerID=8YFLogxK

U2 - 10.1103/PhysRevX.9.011003

DO - 10.1103/PhysRevX.9.011003

M3 - Article

AN - SCOPUS:85062007999

VL - 9

JO - Physical Review X

JF - Physical Review X

SN - 2160-3308

IS - 1

M1 - 011003

ER -