arXiv - PHYS - Disordered Systems and Neural Networks最新文献_第8页

Graph Neural Networks for Predicting Solubility in Diverse Solvents using MolMerger incorporating Solute-solvent Interactions 利用包含溶质-溶剂相互作用的 MolMerger 图神经网络预测不同溶剂中的溶解度

arXiv - PHYS - Disordered Systems and Neural Networks Pub Date : 2024-02-17 DOI: arxiv-2402.11340

Vansh Ramani, Tarak Karmakar

{"title":"Graph Neural Networks for Predicting Solubility in Diverse Solvents using MolMerger incorporating Solute-solvent Interactions","authors":"Vansh Ramani, Tarak Karmakar","doi":"arxiv-2402.11340","DOIUrl":"https://doi.org/arxiv-2402.11340","url":null,"abstract":"Prediction of solubility has been a complex and challenging physiochemical\u0000problem that has tremendous implications in the chemical and pharmaceutical\u0000industry. Recent advancements in machine learning methods have provided great\u0000scope for predicting the reliable solubility of a large number of molecular\u0000systems. However, most of these methods rely on using physical properties\u0000obtained from experiments and or expensive quantum chemical calculations. Here,\u0000we developed a method that utilizes a graphical representation of\u0000solute-solvent interactions using `MolMerger', which captures the strongest\u0000polar interactions between molecules using Gasteiger charges and creates a\u0000graph incorporating the true nature of the system. Using these graphs as input,\u0000a neural network learns the correlation between the structural properties of a\u0000molecule in the form of node embedding and its physiochemical properties as\u0000output. This approach has been used to calculate molecular solubility by\u0000predicting the Log solubility values of various organic molecules and\u0000pharmaceuticals in diverse sets of solvents.","PeriodicalId":501066,"journal":{"name":"arXiv - PHYS - Disordered Systems and Neural Networks","volume":"27 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139920722","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Random features and polynomial rules 随机特征和多项式规则

arXiv - PHYS - Disordered Systems and Neural Networks Pub Date : 2024-02-15 DOI: arxiv-2402.10164

Fabián Aguirre-López, Silvio Franz, Mauro Pastore

{"title":"Random features and polynomial rules","authors":"Fabián Aguirre-López, Silvio Franz, Mauro Pastore","doi":"arxiv-2402.10164","DOIUrl":"https://doi.org/arxiv-2402.10164","url":null,"abstract":"Random features models play a distinguished role in the theory of deep\u0000learning, describing the behavior of neural networks close to their\u0000infinite-width limit. In this work, we present a thorough analysis of the\u0000generalization performance of random features models for generic supervised\u0000learning problems with Gaussian data. Our approach, built with tools from the\u0000statistical mechanics of disordered systems, maps the random features model to\u0000an equivalent polynomial model, and allows us to plot average generalization\u0000curves as functions of the two main control parameters of the problem: the\u0000number of random features $N$ and the size $P$ of the training set, both\u0000assumed to scale as powers in the input dimension $D$. Our results extend the\u0000case of proportional scaling between $N$, $P$ and $D$. They are in accordance\u0000with rigorous bounds known for certain particular learning tasks and are in\u0000quantitative agreement with numerical experiments performed over many order of\u0000magnitudes of $N$ and $P$. We find good agreement also far from the asymptotic\u0000limits where $Dto infty$ and at least one between $P/D^K$, $N/D^L$ remains\u0000finite.","PeriodicalId":501066,"journal":{"name":"arXiv - PHYS - Disordered Systems and Neural Networks","volume":"99 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139755558","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Designing three-way entangled and nonlocal two-way entangled single particle states via alternate quantum walks 通过交替量子行走设计三向纠缠和非局域双向纠缠单粒子态

arXiv - PHYS - Disordered Systems and Neural Networks Pub Date : 2024-02-07 DOI: arxiv-2402.05080

Dinesh Kumar Panda, Colin Benjamin

引用次数: 0

Asymptotic generalization error of a single-layer graph convolutional network 单层图卷积网络的渐近泛化误差

arXiv - PHYS - Disordered Systems and Neural Networks Pub Date : 2024-02-06 DOI: arxiv-2402.03818

O. Duranthon, L. Zdeborová

引用次数: 0

Nature-Inspired Local Propagation 自然启发的本地传播

arXiv - PHYS - Disordered Systems and Neural Networks Pub Date : 2024-02-04 DOI: arxiv-2402.05959

Alessandro Betti, Marco Gori

引用次数: 0

Why is the Dynamics of Glasses Super-Arrhenius? 玻璃的动力学为什么是超阿伦尼乌斯？

arXiv - PHYS - Disordered Systems and Neural Networks Pub Date : 2024-02-02 DOI: arxiv-2402.01883

Jean-Philippe BouchaudCapital Fund Management & Académie des Sciences

{"title":"Why is the Dynamics of Glasses Super-Arrhenius?","authors":"Jean-Philippe BouchaudCapital Fund Management & Académie des Sciences","doi":"arxiv-2402.01883","DOIUrl":"https://doi.org/arxiv-2402.01883","url":null,"abstract":"The steep increase of the relaxation time of glass forming liquids upon\u0000cooling is traditionally ascribed to an impending entropy crisis: since the\u0000system has \"nowhere to go\", dynamics must come to a halt. This classic\u0000argument, due to Adam & Gibbs, has been bolstered and refined by the\u0000development of the Random First Order Transition (RFOT) theory, which fares\u0000remarkably well at reproducing most salient experimental facts of super-cooled\u0000liquids. All static predictions of RFOT, including the existence of a\u0000point-to-set length and the role of pinning sites, have been vindicated by\u0000detailed numerical simulations. Yet, there is no consensus that the basic\u0000mechanism explaining the glass transition is the one captured by RFOT. Strong\u0000doubts have emerged following the observation that adding or removing kinetic\u0000constraints can change the relaxation time by orders of magnitude, while\u0000leaving thermodynamics unchanged. This is at odds with the idea of a one-to-one\u0000mapping between excess entropy and relaxation time. In the following discussion paper presented at the Solvay conference in\u0000October 2023, we review areas of consensus and dissent of RFOT with other\u0000competing theoretical proposals, and propose possible paths for (partial)\u0000reconciliation. We further argue that extensive numerical simulations of the\u0000non-linear susceptibility of glasses, in particular in the aging regime, should\u0000shed important light on the mechanism at the origin of the super-Arrhenius\u0000behaviour of the relaxation time. In any case, more imagination is still needed\u0000to come up with experimental, theoretical or numerical ideas that would allow\u0000to finally settle the question of why glasses do not flow.","PeriodicalId":501066,"journal":{"name":"arXiv - PHYS - Disordered Systems and Neural Networks","volume":"7 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139755411","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Catching thermal avalanches in disordered XXZ model 在无序 XXZ 模型中捕捉热雪崩

arXiv - PHYS - Disordered Systems and Neural Networks Pub Date : 2024-02-02 DOI: arxiv-2402.01362

Tomasz Szołdra, Piotr Sierant, Maciej Lewenstein, Jakub Zakrzewski

{"title":"Catching thermal avalanches in disordered XXZ model","authors":"Tomasz Szołdra, Piotr Sierant, Maciej Lewenstein, Jakub Zakrzewski","doi":"arxiv-2402.01362","DOIUrl":"https://doi.org/arxiv-2402.01362","url":null,"abstract":"We study the XXZ model with a random magnetic field in contact with a weakly\u0000disordered spin chain, acting as a finite thermal bath. We revise Fermi's\u0000golden rule description of the interaction between the thermal bath and the XXZ\u0000spin chain, contrasting it with a non-perturbative quantum avalanche scenario\u0000for the thermalization of the system. We employ two-point correlation functions\u0000to define the extent, $xi_d$, of the thermalized region next to the bath.\u0000Unbounded growth of $xi_d$ proportional to the logarithm of time or faster is\u0000a signature of an avalanche. It signifies the thermalization of the system, as\u0000we confirm numerically for a generic initial state in the ergodic and critical\u0000regimes of the XXZ spin chain. In the many-body localized regime, a clear\u0000termination of avalanches is observed for specifically prepared initial states\u0000and, surprisingly, is not visible for generic initial product states.\u0000Additionally, we extract the localization length of the local integrals of\u0000motion and show that a bath made out of a weakly disordered XXZ chain has a\u0000similar effect on the system as a bath modeled by a Hamiltonian from a Gaussian\u0000Orthogonal Ensemble of random matrices. We also comment on the result of the\u0000earlier study (Phys. Rev. B 108, L020201 (2023)), arguing that the observed\u0000thermalization is due to external driving of the system and does not occur in\u0000the autonomous model. Our work reveals experimentally accessible signatures of\u0000quantum avalanches and identifies conditions under which termination of the\u0000avalanches may be observed.","PeriodicalId":501066,"journal":{"name":"arXiv - PHYS - Disordered Systems and Neural Networks","volume":"38 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139689445","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Onset of nonequilibrium in a driven Anderson insulator 驱动型安德森绝缘体的非平衡起始

arXiv - PHYS - Disordered Systems and Neural Networks Pub Date : 2024-01-30 DOI: arxiv-2401.16940

Z. Ovadyahu

引用次数: 0

Anomalous localization in spin-chain with tilted interactions 具有倾斜相互作用的自旋链中的反常定位

arXiv - PHYS - Disordered Systems and Neural Networks Pub Date : 2024-01-25 DOI: arxiv-2401.14369

Arindam Mallick, Jakub Zakrzewski

引用次数: 0

Localization of Lindbladian Fermions 林德布拉第费米子的定位

arXiv - PHYS - Disordered Systems and Neural Networks Pub Date : 2024-01-25 DOI: arxiv-2401.14006

Foster Thompson, Yi Huang, Alex Kamenev

引用次数: 0