4.6 Article

Optimization landscape in the simplest constrained random least-square problem

相关参考文献

注意:仅列出部分参考文献,下载原文获取全部文献信息。
Article Physics, Mathematical

Triviality of the Geometry of Mixed p-Spin Spherical Hamiltonians with External Field

David Belius et al.

Summary: In this study, we investigate isotropic Gaussian random fields on the high-dimensional sphere, with an additional deterministic linear term called mixed p-spin Hamiltonians with external field. It is proven that if the external field is sufficiently strong, the resulting function has trivial geometry, characterized by only two critical points. This is in contrast to the case of no or weak external field, where these functions typically have an exponential number of critical points. The study provides an explicit threshold for the magnitude of the external field, referred to as h(c), which is necessary for the trivialization. It is conjectured that h(c) is sharp. The Kac-Rice formula is utilized as the main tool for analysis. This work extends the findings of Fyodorov [14], which focused on the trivial regime for the special case of pure p-spin Hamiltonians with random external field.

JOURNAL OF STATISTICAL PHYSICS (2022)

Article Physics, Mathematical

Spherical Spin Glass Model with External Field

Jinho Baik et al.

Summary: This study focused on the free energy and overlaps in the 2-spin spherical Sherrington Kirkpatrick spin glass model with an external field, comparing it to the model without an external field. The research involved computing limiting values and fluctuations of the free energy, as well as three types of overlaps, as the strength of the external field approaches zero with increasing dimension of the spin variable. The findings were further supported by Landon and Sosoe in a recent, independent paper.

JOURNAL OF STATISTICAL PHYSICS (2021)

Article Mechanics

The loss surfaces of neural networks with general activation functions

Nicholas P. Baskerville et al.

Summary: The authors tested the validity of previous assumptions about the relationship between training loss surfaces of deep neural networks and spherical multi-spin glass models by removing the restriction on ReLU activation functions. By using supersymmetric methods in random matrix theory, they opened up a new path in spin glass complexity calculations that may be useful in other contexts. Their results shed new light on both the strengths and weaknesses of spin glass models in this specific context.

JOURNAL OF STATISTICAL MECHANICS-THEORY AND EXPERIMENT (2021)

Article Physics, Multidisciplinary

COUNTING STATIONARY POINTS OF THE LOSS FUNCTION IN THE SIMPLEST CONSTRAINED LEAST-SQUARE OPTIMIZATION

Yan Fyodorov et al.

ACTA PHYSICA POLONICA B (2020)

Article Mathematics, Applied

The Landscape of the Spiked Tensor Model

Gerard Ben Arous et al.

COMMUNICATIONS ON PURE AND APPLIED MATHEMATICS (2019)

Article Physics, Mathematical

A Spin Glass Model for Reconstructing Nonlinearly Encrypted Signals Corrupted by Noise

Yan V. Fyodorov

JOURNAL OF STATISTICAL PHYSICS (2019)

Article Mechanics

Topology trivialization transition in random non-gradient autonomous ODEs on a sphere

Y. V. Fyodorov

JOURNAL OF STATISTICAL MECHANICS-THEORY AND EXPERIMENT (2016)

Article Physics, Mathematical

Matrix Optimization Under Random External Fields

Amir Dembo et al.

JOURNAL OF STATISTICAL PHYSICS (2015)

Article Physics, Mathematical

Topology Trivialization and Large Deviations for the Minimum in the Simplest Random Optimization

Yan V. Fyodorov et al.

JOURNAL OF STATISTICAL PHYSICS (2014)

Article Mathematics, Applied

Random matrices and complexity of spin glasses

Antonio Auffinger et al.

COMMUNICATIONS ON PURE AND APPLIED MATHEMATICS (2013)

Article Physics, Multidisciplinary

The averaged characteristic polynomial for the Gaussian and chiral Gaussian ensembles with a source

Peter J. Forrester

JOURNAL OF PHYSICS A-MATHEMATICAL AND THEORETICAL (2013)

Article Physics, Multidisciplinary

Large deviation eigenvalue density for the soft edge Laguerre and Jacobi β-ensembles

Peter J. Forrester

JOURNAL OF PHYSICS A-MATHEMATICAL AND THEORETICAL (2012)

Article Physics, Multidisciplinary

Multifractality and freezing phenomena in random energy landscapes: An introduction

Yan V. Fyodorov

PHYSICA A-STATISTICAL MECHANICS AND ITS APPLICATIONS (2010)

Article Physics, Mathematical

Replica symmetry breaking condition exposed by random matrix calculation of landscape complexity

Yan V. Fyodorov et al.

JOURNAL OF STATISTICAL PHYSICS (2007)