Computer Physics Communications最新文献

{"title":"A field-aligned gyrokinetic solver based on discontinuous Galerkin in tokamak geometry","authors":"Gahyung Jo ,&nbsp;Janghoon Seo ,&nbsp;Jae-Min Kwon ,&nbsp;Eisung Yoon","doi":"10.1016/j.cpc.2025.109769","DOIUrl":"10.1016/j.cpc.2025.109769","url":null,"abstract":"<div><div>This paper presents the development of a hyperbolic solver for the gyrokinetic equation in tokamak geometry. The discontinuous Galerkin method discretizes the gyrokinetic equation on the field-aligned mesh composed of twisted prism-shaped elements in the tokamak domain. The elements are generated by extending the vertices of unstructured triangular elements on a poloidal plane following the equilibrium magnetic field lines. A sub-triangulation is employed to transfer information between nonconforming meshes, which is inevitable when implementing the field-aligned mesh. The numerical integrations of elements in the field-aligned mesh are performed by transforming the numerical integrations of reference elements in a reference element. We investigate the impact of field-aligned mesh on the numerical interpolation of synthetic plasma fluctuation data generated by a ballooning function. The numerical tests show that the field-aligned mesh can significantly improve computational efficiencies. Additionally, we estimate a sufficient condition for a stable temporal discretization of the hyperbolic solver based on a Runge-Kutta method. The estimation indicates that the field-aligned mesh can allow a notable increase of the time step size for stable simulation. In the numerical experiments, the solver shows good conservations of physical quantities such as mass, kinetic energy, and toroidal canonical angular momentum.</div></div>","PeriodicalId":285,"journal":{"name":"Computer Physics Communications","volume":"316 ","pages":"Article 109769"},"PeriodicalIF":7.2,"publicationDate":"2025-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144703530","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"PARPHOM: PARallel PHOnon calculator for Moiré systems","authors":"Shinjan Mandal ,&nbsp;Indrajit Maity ,&nbsp;H.R. Krishnamurthy ,&nbsp;Manish Jain","doi":"10.1016/j.cpc.2025.109760","DOIUrl":"10.1016/j.cpc.2025.109760","url":null,"abstract":"<div><div>The introduction of a twist between two layers of two-dimensional materials has opened up a new and exciting field of research known as twistronics. In these systems, the phonon dispersions show significant renormalization and enhanced electron-phonon interactions as a function of the twist angle. However, the large system size of the resulting moiré patterns in these systems makes phonon calculations computationally challenging. In this paper, we present PARPHOM, a powerful code package designed to address these challenges. PARPHOM enables the generation of force constants, computation of phononic band structures, and determination of density of states in twisted 2D material systems. Moreover, PARPHOM provides essential routines to investigate the finite temperature dynamics in these systems and analyze the chirality of the phonon bands. This paper serves as an introduction to PARPHOM, highlighting its capabilities and demonstrating its utility in unraveling the intricate phononic properties of twisted 2D materials.</div></div><div><h3>Program summary</h3><div><em>Program Title:</em> PARPHOM</div><div><em>CPC Library link to program files:</em> <span><span>https://doi.org/10.17632/gp6rzrp47m.1</span><svg><path></path></svg></span></div><div><em>Developer's repository link:</em> <span><span>https://github.com/qtm-iisc/PARPHOM</span><svg><path></path></svg></span></div><div><em>Licensing provisions:</em> GNU General Public License v3.0</div><div><em>Programming language:</em> FORTRAN, Python</div><div><em>External Routines/libraries:</em> numpy, LAMMPS (serial Python wrapper), mpi4py, ScaLAPACK, HDF5, matplotlib, scipy, Spglib</div><div><em>Nature of problem:</em> Due to the large number of atoms in 2D moiré systems, performing phonon calculations is quite challenging. The exorbitantly high memory requirements of such calculations make them infeasible with currently available codes.</div><div><em>Solution method:</em> A parallel algorithm to generate the force constant matrices for these large moiré systems has been implemented. Parallel diagonalization routines available in ScaLAPACK are then used to diagonalize the dynamical matrices constructed from the force constants at each <strong>q</strong> points.</div></div>","PeriodicalId":285,"journal":{"name":"Computer Physics Communications","volume":"316 ","pages":"Article 109760"},"PeriodicalIF":7.2,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144695043","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Study of simplified conservation flux scheme for gas kinetics based on OpenFOAM framework II: Rykov model","authors":"Mengbo Zhu ,&nbsp;Qingdian Zhang ,&nbsp;Rui Zhang ,&nbsp;Congshan Zhuo ,&nbsp;Sha Liu ,&nbsp;Chengwen Zhong","doi":"10.1016/j.cpc.2025.109763","DOIUrl":"10.1016/j.cpc.2025.109763","url":null,"abstract":"<div><div>We present a computational fluid dynamics solver for diatomic gases, meticulously developed within the dugksFOAM framework. This solver is built upon a conservative gas kinetic scheme with simplified interface flux evaluations, enabling efficient and accurate solutions of the Rykov model equation. An unstructured discrete velocity space is introduced, in which the velocity points are strategically distributed to balance computational efficiency and numerical accuracy. A sophisticated hybrid parallelization strategy, referred to as X-space parallelization, has also been introduced. It integrates domain decomposition in both physical and velocity spaces, significantly enhancing computational efficiency in large-scale simulations. We further compare the computational efficiency between the structured and unstructured velocity space approaches, demonstrating that the unstructured configuration achieves notable reductions in computational cost without compromising accuracy. Moreover, the parallel performance of the solver is systematically evaluated under both small- and large-scale settings, showcasing excellent scalability and robustness. The accuracy and reliability of the solver are validated against a comprehensive set of benchmark cases, including shock structure problems, lid-driven cavity flow, supersonic flows past a flat plate, cylindrical blunt body, and sphere. These results convincingly confirm the solver's capability to capture a wide range of rarefied flow phenomena in diatomic gases, from one-dimensional to three-dimensional flows.</div></div>","PeriodicalId":285,"journal":{"name":"Computer Physics Communications","volume":"316 ","pages":"Article 109763"},"PeriodicalIF":7.2,"publicationDate":"2025-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144695040","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Performance optimization of GJK collision detection in discrete element simulations","authors":"Alireza Yazdani ,&nbsp;Anthony Wachs","doi":"10.1016/j.cpc.2025.109768","DOIUrl":"10.1016/j.cpc.2025.109768","url":null,"abstract":"<div><div>This paper presents a comprehensive performance analysis of the Gilbert-Johnson-Keerthi (GJK) algorithm and its variants in the context of Discrete Element Method (DEM) simulations. Various optimization techniques, including bounding volumes, different distance sub-algorithms, Nesterov acceleration, and temporal coherence are investigated to evaluate their impact on computational efficiency for different particle shapes and aspect ratios. The study considers both static packing and rotating drum benchmarks, covering a wide range of particle geometries such as cubes, icosahedrons, cylinders, and superquadrics. Our findings indicate that the choice of bounding volume technique significantly affects performance, with oriented bounding cylinder outperforming oriented bounding boxes for elongated particles. Nesterov acceleration, although theoretically promising, generally shows limited performance improvements except for highly spherical particles. Temporal coherence, while beneficial for certain particle shapes and moderate aspect ratios, is less effective when particles are highly elongated or distant from each other. These results offer valuable insights for optimizing DEM simulations involving complex particle shapes and varying elongation levels.</div></div>","PeriodicalId":285,"journal":{"name":"Computer Physics Communications","volume":"316 ","pages":"Article 109768"},"PeriodicalIF":7.2,"publicationDate":"2025-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144695025","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Dynasor 2: From simulation to experiment through correlation functions","authors":"Esmée Berger ,&nbsp;Erik Fransson ,&nbsp;Fredrik Eriksson ,&nbsp;Eric Lindgren ,&nbsp;Göran Wahnström ,&nbsp;Thomas Holm Rod ,&nbsp;Paul Erhart","doi":"10.1016/j.cpc.2025.109759","DOIUrl":"10.1016/j.cpc.2025.109759","url":null,"abstract":"<div><div>Correlation functions, such as static and dynamic structure factors, offer a versatile approach to analyzing atomic-scale structure and dynamics. By having access to the full dynamics from atomistic simulations, they serve as valuable tools for understanding material behavior. Experimentally, material properties are commonly probed through scattering measurements, which also provide access to static and dynamic structure factors. However, it is not trivial to decode these due to complex interactions between atomic motion and the probe. Atomistic simulations can help bridge this gap, allowing for detailed understanding of the underlying dynamics. In this paper, we illustrate how correlation functions provide structural and dynamical insights from simulation and showcase the strong agreement with experiment. To compute the correlation functions, we have updated the Python package <span>dynasor</span> with a new interface and, importantly, added support for weighting the computed quantities with form factors or cross sections, facilitating direct comparison with probe-specific structure factors. Additionally, we have incorporated the spectral energy density method, which offers an alternative view of the dispersion for crystalline systems, as well as functionality to project atomic dynamics onto phonon modes, enabling detailed analysis of specific phonon modes from atomistic simulation. We illustrate the capabilities of <span>dynasor</span> with diverse examples, ranging from liquid <figure><img></figure> to perovskites, and compare computed results with X-ray, electron and neutron scattering experiments. This highlights how computed correlation functions can not only agree well with experimental observations, but also provide deeper insight into the atomic-scale structure and dynamics of a material.</div></div>","PeriodicalId":285,"journal":{"name":"Computer Physics Communications","volume":"316 ","pages":"Article 109759"},"PeriodicalIF":7.2,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144695041","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Intelligent cross-linking in polymer simulations: SuSi’s approach to complex 3D networks","authors":"David Naranjo ,&nbsp;Didac Martí ,&nbsp;Carlos Alemán ,&nbsp;José García-Torres ,&nbsp;Juan Torras","doi":"10.1016/j.cpc.2025.109767","DOIUrl":"10.1016/j.cpc.2025.109767","url":null,"abstract":"<div><div>Cross-linked polymers play a vital role in the materials science due to their mechanical strength, chemical resistance, and thermal stability, making them invaluable in biomedical devices, coatings, and electronics. However, constructing realistic molecular models of these systems remains a challenge due to their complex cross-linked networks. This study introduces SuSi, a Python-based program designed to generate both linear and cross-linked polymer systems for molecular simulations. SuSi uses artificial intelligence tree search algorithms to optimize the cross-linking process, ensuring efficient and collision-free network formation. The program is compatible with the AMBER force field and supports a wide variety of polymer architectures, including homopolymers, block copolymers, and complex 3D-network structures. To demonstrate its capabilities, SuSi was employed to generate three distinct cross-linked systems: silane-cross-linked polyethylene (Si-XLPE), thermosensitive poly(NIPAAm-co-MBA), and the complex unsaturated polyesteramide hydrogel made of phenylalanine, butenediol, and fumarate, cross-linked with polyethylene glycol (UPEA-PEG). The generated structures were successfully parametrized for molecular dynamics simulations and validated through experimental observables, showing that SuSi is a versatile tool for accurately modeling complex polymeric systems and advancing polymer simulations.</div></div>","PeriodicalId":285,"journal":{"name":"Computer Physics Communications","volume":"316 ","pages":"Article 109767"},"PeriodicalIF":7.2,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144695042","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CWebGen A tool to study colour structure of scattering amplitudes in IR limit","authors":"Neelima Agarwal ,&nbsp;Sourav Pal ,&nbsp;Aditya Srivastav ,&nbsp;Anurag Tripathi","doi":"10.1016/j.cpc.2025.109765","DOIUrl":"10.1016/j.cpc.2025.109765","url":null,"abstract":"<div><div>Infrared singularities in perturbative Quantum Chromodynamics (QCD) are captured by the Soft function, which can be calculated efficiently in terms of multiparton webs. Web is a closed set of diagrams whose colour and kinematics mix through a web mixing matrix. The web mixing matrices are computed using a well known replica trick algorithm. We present a package implemented in Mathematica to calculate these mixing matrices. Along with the package, we provide benchmark points for several state-of-the art computations.</div></div>","PeriodicalId":285,"journal":{"name":"Computer Physics Communications","volume":"316 ","pages":"Article 109765"},"PeriodicalIF":7.2,"publicationDate":"2025-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144679110","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Integrating physics and data-driven approaches: An explainable and uncertainty-aware hybrid model for wind turbine power prediction","authors":"Alfonso Gijón ,&nbsp;Simone Eiraudo ,&nbsp;Antonio Manjavacas ,&nbsp;Daniele Salvatore Schiera ,&nbsp;Miguel Molina-Solana ,&nbsp;Juan Gómez-Romero","doi":"10.1016/j.cpc.2025.109761","DOIUrl":"10.1016/j.cpc.2025.109761","url":null,"abstract":"<div><div>The rapid growth of the wind energy sector underscores the urgent need to optimize turbine operations and ensure effective maintenance through early fault detection systems. While traditional empirical and physics-based models offer approximate predictions of power generation based on wind speed, they often fail to capture the complex, non-linear relationships between other input variables and the resulting power output. Data-driven machine learning methods present a promising avenue for improving wind turbine modeling by leveraging large datasets, enhancing prediction accuracy but often at the cost of interpretability. In this study, we propose a hybrid semi-parametric model that combines the strengths of both approaches, applied to a dataset from a wind farm with four turbines. The model integrates a physics-inspired submodel, providing a reasonable approximation of power generation, with a non-parametric submodel that predicts the residuals. This non-parametric submodel is trained on a broader range of variables to account for phenomena not captured by the physics-based component. The hybrid model achieves a 37% improvement in prediction accuracy over the physics-based model and performs comparably to a purely data-driven reference model, while offering additional advantages in terms of explainability and robustness. To further enhance interpretability, SHAP values are used to analyze the influence of input features on the residual submodel's output. Additionally, prediction uncertainties are quantified using a conformalized quantile regression method. The combination of these techniques, alongside the physics grounding of the parametric submodel, provides a flexible, accurate, and reliable framework. Ultimately, this study opens the door for evaluating the impact of unmodeled phenomena on wind turbine power generation, offering a basis for potential optimization.</div></div>","PeriodicalId":285,"journal":{"name":"Computer Physics Communications","volume":"316 ","pages":"Article 109761"},"PeriodicalIF":7.2,"publicationDate":"2025-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144672337","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Machine learning assisted canonical sampling (Mlacs)","authors":"Aloïs Castellano ,&nbsp;Romuald Béjaud ,&nbsp;Pauline Richard ,&nbsp;Olivier Nadeau ,&nbsp;Clément Duval ,&nbsp;Grégory Geneste ,&nbsp;Gabriel Antonius ,&nbsp;Johann Bouchet ,&nbsp;Antoine Levitt ,&nbsp;Gabriel Stoltz ,&nbsp;François Bottin","doi":"10.1016/j.cpc.2025.109730","DOIUrl":"10.1016/j.cpc.2025.109730","url":null,"abstract":"&lt;div&gt;&lt;div&gt;The acceleration of material property calculations while maintaining &lt;em&gt;ab initio&lt;/em&gt; accuracy (1 meV/atom) is one of the major challenges in computational physics. In this paper, we introduce a Python package enhancing the computation of (finite temperature) material properties at the &lt;em&gt;ab initio&lt;/em&gt; level using machine learning interatomic potentials (MLIP). The Machine Learning Assisted Canonical Sampling (&lt;span&gt;Mlacs&lt;/span&gt;) method, grounded in a self-consistent variational approach, iteratively trains a MLIP using an active learning strategy in order to significantly reduce the computational cost of &lt;em&gt;ab initio&lt;/em&gt; simulations.&lt;/div&gt;&lt;div&gt;&lt;span&gt;Mlacs&lt;/span&gt; offers a modular and user-friendly interface that seamlessly integrates Density Functional Theory (DFT) codes, MLIP potentials, and molecular dynamics packages, enabling a wide range of applications, while maintaining a near-DFT accuracy. These include sampling the canonical ensemble of a system, performing free energy calculations, transition path sampling, and geometry optimization, all by utilizing surrogate MLIP potentials, in place of &lt;em&gt;ab initio&lt;/em&gt; calculations.&lt;/div&gt;&lt;div&gt;This paper provides a comprehensive overview of the theoretical foundations and implementation of the &lt;span&gt;Mlacs&lt;/span&gt; method. We also demonstrate its accuracy and efficiency through various examples, showcasing the capabilities of the &lt;span&gt;Mlacs&lt;/span&gt; package.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Program summary&lt;/h3&gt;&lt;div&gt;&lt;em&gt;Program title:&lt;/em&gt; &lt;span&gt;Mlacs&lt;/span&gt;&lt;/div&gt;&lt;div&gt;&lt;em&gt;CPC Library link to program files:&lt;/em&gt; &lt;span&gt;&lt;span&gt;https://doi.org/10.17632/vtfzjnc6cr.1&lt;/span&gt;&lt;svg&gt;&lt;path&gt;&lt;/path&gt;&lt;/svg&gt;&lt;/span&gt;&lt;/div&gt;&lt;div&gt;&lt;em&gt;Licensing provisions:&lt;/em&gt; GNU General Public License, version 3&lt;/div&gt;&lt;div&gt;&lt;em&gt;Programming language:&lt;/em&gt; Python&lt;/div&gt;&lt;div&gt;&lt;em&gt;Nature of problem:&lt;/em&gt; Numerous material properties, whether related to the ground state or finite temperature thermodynamic quantities, cannot be deduced from classical simulations and require accurate but highly demanding &lt;em&gt;ab initio&lt;/em&gt; calculations. Enhancing the efficiency of these simulations while preserving a near-&lt;em&gt;ab initio&lt;/em&gt; accuracy is one of the biggest challenges in modern computational physics.&lt;/div&gt;&lt;div&gt;&lt;em&gt;Solution method:&lt;/em&gt; The emergence of MLIP potentials enables us to tackle this issue. The method implemented in &lt;span&gt;Mlacs&lt;/span&gt; allows for the acceleration of &lt;em&gt;ab initio&lt;/em&gt; calculations by training a MLIP potential on the fly. At the end of the simulation, &lt;span&gt;Mlacs&lt;/span&gt; produces an optimal local surrogate potential, a database that includes a sample of representative atomic configurations with their statistical weights, as well as information on convergence control and thermodynamic quantities.&lt;/div&gt;&lt;div&gt;&lt;em&gt;Additional comments:&lt;/em&gt; The seminal version is defined in [1]. The new version [2], &lt;span&gt;Mlacs&lt;/span&gt; v1.0.2, works on various architectures and includes several new features.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;References&lt;/h3&gt;&lt;div&gt;&lt;ul&gt;","PeriodicalId":285,"journal":{"name":"Computer Physics Communications","volume":"316 ","pages":"Article 109730"},"PeriodicalIF":7.2,"publicationDate":"2025-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144679111","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Generative flow induced neural architecture search: Towards discovering optimal architecture in wavelet neural operator","authors":"Hartej Soin ,&nbsp;Tapas Tripura ,&nbsp;Souvik Chakraborty","doi":"10.1016/j.cpc.2025.109755","DOIUrl":"10.1016/j.cpc.2025.109755","url":null,"abstract":"<div><div>We propose a generative flow-induced neural architecture search algorithm. The proposed approach devises simple feed-forward neural networks to learn stochastic policies to generate sequences of architecture hyperparameters such that the generated states are in proportion to the reward from the terminal state. We demonstrate the efficacy of the proposed search algorithm on the wavelet neural operator (WNO), where we learn a policy to generate a sequence of hyperparameters like wavelet basis and activation operators for wavelet integral blocks. While the trajectory of the generated wavelet basis and activation sequence is cast as flow, the policy is learned by minimizing the flow violation between each state in the trajectory and maximizing the reward from the terminal state. In the terminal state, we train WNO simultaneously to guide the search. We propose using the negative exponent of the WNO loss on the validation dataset as the reward function. While the grid search-based neural architecture generation algorithms foresee every combination, the proposed framework generates the most probable sequence based on the positive reward from the terminal state, thereby reducing exploration time. Compared to reinforcement learning schemes, where complete episodic training is required to get the reward, the proposed algorithm generates the hyperparameter trajectory sequentially. Through four fluid mechanics-oriented problems, we illustrate that the learned policies can sample the best-performing architecture of the neural operator, thereby improving the performance of the vanilla wavelet neural operator. We compare the performance of the proposed flow-based search strategy with that of a Monte Carlo Tree Search (MCTS) -based algorithm and observe an improvement of ≥23% in the resulting optimal architecture.</div></div>","PeriodicalId":285,"journal":{"name":"Computer Physics Communications","volume":"316 ","pages":"Article 109755"},"PeriodicalIF":7.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144654901","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}