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 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.

arXiv commentv2 with references added, typos corrected

References in corpus (28)

• Spin-Glass Theory for Pedestrians
• The structural glass transition and the entropy of the metastable states
• Statistical physics of inference: Thresholds and algorithms
• Replica Field Theory for Deterministic Models (II): A Non-Random Spin Glass with Glassy Behavior
• Ideal Glass Transitions by Random Pinning
• On the stationary points of the TAP free energy
• Transition to chaos in random neuronal networks
• The statistics of critical points of Gaussian fields on large-dimensional spaces
• Asymptotic power of sphericity tests for high-dimensional data
• Topological and Dynamical Complexity of Random Neural Networks
• Complexity of random smooth functions on the high-dimensional sphere
• Nonlinear Analogue of the May-Wigner Instability Transition
• Critical Behaviour of the Number of Minima of a Random Landscape at the Glass Transition Point and the Tracy-Widom distribution
• The Complexity of Ising Spin Glasses
• Topology trivialization and large deviations for the minimum in the simplest random optimization
• Parisi formula, disorder chaos and fluctuation for the ground state energy in the spherical mixed p-spin models
• Complexity in Mean-Field Spin-Glass Models: Ising $p$-spin
• Quenched complexity of the p-spin spherical spin-glass with external magnetic field
• Low Temperature Asymptotics in Spherical Mean Field Spin Glasses
• Marginal States in Mean Field Glasses
• On Spin-Glass Complexity
• Topology trivialization transition in random non-gradient autonomous ODE's on a sphere
• Free energy landscapes, dynamics and the edge of chaos in mean-field models of spin glasses
• Coexistence of supersymmetric and supersymmetry-breaking states in spherical spin-glasses
• On Supersymmetry Breaking in the Computation of the Complexity
• Exactly solvable spin-glass models with ferromagnetic couplings: the spherical multi-$p$-spin model in a self-induced field
• p>2 spin glasses with first order ferromagnetic transitions
• Tap Complexity, the Cavity Method and Supersymmetry

Cited by in corpus (40)

• Algorithmic thresholds for tensor PCA
• Generalized Lotka-Volterra equations with random, non-reciprocal interactions: the typical number of equilibria
• Relaxation dynamics in the energy landscape of glass-forming liquids
• Marvels and Pitfalls of the Langevin Algorithm in Noisy High-dimensional Inference
• Gradient descent dynamics in the mixed $p$-spin spherical model: finite size simulation and comparison with mean-field integration
• How to iron out rough landscapes and get optimal performances: Averaged Gradient Descent and its application to tensor PCA
• Path Integral Approach Unveils the Role of Complex Energy Landscape for Activated Dynamics of Glassy Systems
• How to count in hierarchical landscapes: a 'full' solution to mean-field complexity
• The Loss Surfaces of Neural Networks with General Activation Functions
• Counting equilibria in a random non-gradient dynamics with heterogeneous relaxation rates
• Nonlinearity-generated Resilience in Large Complex Systems
• Quenched complexity of equilibria for asymmetric Generalized Lotka-Volterra equations
• Distribution of rare saddles in the $p$-spin energy landscape
• Barriers, trapping times and overlaps between local minima in the dynamics of the disordered Ising $p$-spin Model
• Manifolds in high dimensional random landscape: complexity of stationary points and depinning
• Phase transition in random tensors with multiple independent spikes
• A spin-glass model for the loss surfaces of generative adversarial networks
• Jamming, relaxation, and memory in a structureless glass former
• Low energy excitations of mean-field glasses
• Manipulating the Glass Transition in Nanoscale
• Discovery of a paired Gaussian and long-tailed distribution of potential energies in nanoglasses
• Thresholds of descending algorithms in inference problems
• Sharp complexity asymptotics and topological trivialization for the (p, k) spiked tensor model
• Competition between Barrier- and Entropy-Driven Activation in Glasses
• The Distribution of Vacua in Random Landscape Potentials
• Manifolds pinned by a high-dimensional random landscape: Hessian at the global energy minimum
• Universal characteristics of deep neural network loss surfaces from random matrix theory
• A constrained TAP approach for disordered spin models: application to the mixed spherical case
• Right large deviation principle for the top eigenvalue of the sum or product of invariant random matrices
• Superposition of Random Plane Waves in High Spatial Dimensions: Random Matrix Approach to Landscape Complexity
• Dipolar Ising Model: Phases, Growth Laws and Universality
• Recursive Algorithm to the Centroid of Free Area for Inherent Structure and Hopping Motion in Deeply Supercooled Binary Hard Disk Systems
• A path integral approach to sparse non-Hermitian random matrices
• Generalized Approximate Survey Propagation for High-Dimensional Estimation
• Far-from-equilibrium complex landscapes
• Functional Renormalization Group Approach for Signal Detection
• Rare Trajectories in a Prototypical Mean-field Disordered Model: Insights into Landscape and Instantons
• Optimization landscape in the simplest constrained random least-square problem
• Landscape Complexity for the Empirical Risk of Generalized Linear Models: Discrimination between Structured Data
• Kinetic glass transition in granular gases and non-linear molecular fluids