arXiv Science⌕ Search
Home/Archive/arXiv

arXiv · 2610.10096

Revealing more on complex energy landscapes by passing less local information: Cavity approach with trust region in non-convex optimization problems

Zhiqi Xu·Chi Ho Yeung
View original source ↗Read source document ↗Export metadata ↓

Abstract

Energy landscapes of non-convex optimization problems are high-dimensional surfaces that are difficult to reveal, visualize, or analyze. Message-passing algorithms, or equivalently cavity approaches in statistical physics, may fail to identify local minima as messages do not converge owing to the ruggedness of the energy landscapes. Here we aim to reveal the characteristics of these complex energy landscapes by limiting the amount of information passed in local messages, improving the convergence of messages for local minima. By studying a routing optimization problem with its convexity governed by a single parameter, we introduce a cavity message-passing algorithm with a trust region, and sample ensembles of converged local minima of rugged energy landscapes across many independent realizations of the same problem instances. We observe three regimes with different characteristics of the energy landscapes: (1) a maximally rugged regime with a peak in the diversity of low-energy minima; (2) an intermediate regime with a hierarchical multi-cluster organization of low-energy solutions; and (3) a smooth regime dominated by a single minimum. Additional tests show that the converged energies are largely insensitive to the size of the trust region if the size is small or moderate, demonstrating the robustness of the proposed approach in identifying characteristics of rugged energy landscapes.

Explore related subjects

cond-mat.dis-nncond-mat.stat-mechphysics.soc-ph

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Zhiqi Xu, Chi Ho Yeung. 2026-10-07. Revealing more on complex energy landscapes by passing less local information: Cavity approach with trust region in non-convex optimization problems. https://arxiv.org/abs/2610.10096

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Exact distribution of the output of a deep-layered machine

Deep learning builds complex global functions through the repeated composition of local functions. But the resulting distribution over global functions is unknown. We derive this distribution exactly for finite width and depth when the local functions are chosen at random. Surprisingly, depth first creates a sharply structured distribution, then erases this structure as probability drains into two absorbing states. This competition produces a crossover at a depth exponential in the number of inputs, beyond which additional layers mainly drive collapse. The transient structure may help explain why highly overparameterized networks can still generalize.

cond-mat.dis-nn↗
arXiv2026

Dynamic Pseudogap Model

We formulate a theory of pseudogap formation generated by dynamic finite nesting vector ${\bf Q}$ fluctuations with characteristic oscillation frequency $ω_0$, and damping $γ$. Starting from a Hamiltonian describing electrons coupled to a classical Gaussian random field, we introduce the double - series representations of the single - particle Green's function within an Abelian (commuting) approximation to the exact SU(2) time evolution of the pseudogap problem. The resulting propagator naturally acquires a generalized Bogoliubov structure in which every stochastic scattering history is characterized by an effective dynamic gap, leading to a coherent superposition of dynamically broadened sidebands with complex Poisson weights. A central result of the theory is the emergence of a dynamically generated decoherence scale $Γ_{\rm eff}$ governing the crossover between two qualitatively different pseudogap regimes. For $ω_0>Γ_{\rm eff}$ the fluctuating field is resolved coherently and the double-series representation provides a controlled description of dynamic sideband formation. Conversely, when $Γ_{\rm eff}\gtrsimω_0$, coherence is progressively lost and the theory crosses over to the quasistatic fluctuating - gap regime described by the exact continued-fraction solution. The coherent and quasistatic descriptions are therefore interpreted as two complementary asymptotic limits of the same microscopic dynamic pseudogap model. The detailed results of numerical calculations for electron spectral density and density of states supporting our approach are presented for different sets of model parameters confirming the general picture of this crossover.

cond-mat.dis-nn↗
arXiv2026

Predicting activation-barrier and plasticity-onset statistics in a model of glasses

A recently introduced anharmonic mean-field model unifiedly reproduced a broad range of low-temperature glass phenomena --- including harmonic nonphononic spectral properties, linear micromechanics and strongly driven elasto-plastic dynamics --- indicating that its underlying energy landscape is intrinsically glassy. Here, we apply a nonlinear modes framework to the model and derive analytic predictions for the asymptotic distributions of activation barriers $p(Δ{U})\!\sim\!(Δ{U})^{1/4}$ and the external force needed for the onset of plasticity $p(f_{\rm c})\sim f_{\rm c}^{2/3}$, for their extreme-value scaling and for $\langleΔ{U}\rangle$ beyond the asymptotic regime. These predictions are expected to equally apply to the mean-field model and to finite-dimensional glasses. We develop efficient algorithms for sampling minima and saddles of the model's glassy potential energy landscape, and quantitatively confirm the theoretical predictions. This progress is enabled by identifying a subset of collective degrees of freedom that are physically relevant for activated glassy dynamics, which like the theoretical predictions should apply to realistic glasses.

cond-mat.dis-nn↗