[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-125892-en":3,"doc-seo-125892-105":31,"detail-sidebar-cat-0-en-105":93},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":28,"seo_description":14,"update_tm":29,"read_time":30},125892,2336474466712,"Maeve","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","Integration of Physics-Derived Memristor Models with Machine Learning Frameworks - Research Report","Simulation frameworks such as MemTorch, DNN+NeuroSim, and aihwkit enable end-to-end co-design of memristive machine-learning accelerators, but commonly rely on lookup-table or simplified analytic memristor models. This limits fidelity for key nonidealities that stem from physical switching dynamics. The work integrates a physics-derived Valence Change Memory (VCM) model into a GPU-friendly ML simulator, solving an ODE for oxygen-vacancy dynamics. Experiments with MNIST show that noise disrupting SET/RESET matching most strongly impacts network accuracy and can guide device development.","Integration of physics-derived memristor models with machine learning frameworks  \nZhenming Yu∗†, Stephan Menzel∗ , John Paul Strachan∗†, Emre Neftci∗†  \n∗ Fakultt f¨ur Elektrotechnik und Informationstechnik, RWTH Aachen, Aachen, 52074, Germany †Peter Gr¨unberg Institut, Forschungszentrum J¨ulich GmbH, J¨ulich, 52425, Germany  \n{z.yu, [e.neftci](e.neftci}@fz-juelich.de)[}](e.neftci}@fz-juelich.de)[@fz-juelich.de](e.neftci}@fz-juelich.de)  \narXiv :2403 .06746v1 [ cs .ET] 11 Mar 2024  \nAbstract—Simulation frameworks such MemTorch [1] [2], DNN+NeuroSim [3] [4], and aihwkit [5] are commonly used to facilitate the end-to-end co-design of memristive machine learning (ML) accelerators. These simulators can take device nonidealities into account and are integrated with modern ML frameworks. However, memristors in these simulators are modeled with either lookup tables or simple analytic models with basic nonlinearities. These simple models are unable to capture certain performancecritical aspects of device nonidealities. For example, they ignore the physical cause of switching, which induces errors in switching timings and thus incorrect estimations of conductance states. This work aims at bringing physical dynamics into consideration to model nonidealities while being compatible with GPU accelerators. We focus on Valence Change Memory (VCM) cells, where the switching nonlinearity and SET/RESET asymmetry relate tightly with the thermal resistance, ion mobility, Schottky barrier height, parasitic resistance, and other effects [6]. The resulting dynamics require solving an ODE that captures changes in oxygen vacancies. We modified a physics-derived SPICElevel VCM model [7] [8], integrated it with the aihwkit [5] simulator and tested the performance with the MNIST dataset. Results show that noise that disrupts the SET/RESET matching affects network performance the most. This work serves as a tool for evaluating how physical dynamics in memristive devices affect neural network accuracy and can be used to guide the development of future integrated devices.  \nI. INTRODUCTION  \nBecause of their compact size, non-volatility, and low latency, memristive devices show great potential in ML and neuromorphic engineering. Digital, analog, and stochastic inmemory computing schemes have been developed that utilize the advantages of memristors [9] . However, memristors are subject to nonidealities like switching nonlinearity, SET/RESET asymmetry, device-to-device, and cycle-to-cycle variations. Naive training algorithms that don’t take these into account suffer from performance loss. [10] To assist in codesigning memristive ML accelerators, simulation frameworkshave been developed [11] with various memristor models.  \nMemristor models can be generally sorted into two categories: behavioral models and physics-derived models. With behavioral models, memristors are treated as black boxes. Experimental observations are fitted with simple equations, and the models are validated and improved in this process. In contrast, physics-derived models formulate physical equations stemming from an analysis of physical phenomena. The derived equations are often simplified and optimized to get the final solution. While physics-derived models that can accu-  \nrately produce voltage-dependent behaviors have been adopted in circuit design and validations [12] [13], only behavioral models were used in ML simulators like MemTorch [1] [2], DNN+NeuroSim [3] [4], and aihwkit [5] .  \nBehavioral models can produce faithful results at a rather low compute cost, but they are not able to capture some aspects of device physics, which limits the application of ML simulators. For example, they do not model voltage-dependent switching behaviors, so the effect of different SET and RESET voltages cannot be investigated and optimized. They do not model noise based on variations of device parameters, so the simulation results cannot be used to guild material scientists for optimizing memri","cbCaikATVJVHj8tf","https://ap.wps.com/l/cbCaikATVJVHj8tf","pdf",1644034,9,1,5,"English","en",105,"# Abstract\n# Introduction\n# Device Modeling\n## JART VCM Model","[{\"question\":\"Why do existing memristor simulation frameworks struggle to model device nonidealities accurately?\",\"answer\":\"They often use lookup tables or simplified analytic models, which miss physically grounded switching effects such as the cause of switching that affects switching timings and conductance-state estimation.\"},{\"question\":\"What physical dynamics does the proposed work bring into the memristor model?\",\"answer\":\"It incorporates physical dynamics for VCM cells by solving an ODE that captures oxygen-vacancy changes, linking switching nonlinearity and SET/RESET asymmetry to thermal and material-related parameters.\"},{\"question\":\"What is the main result from testing with the MNIST dataset?\",\"answer\":\"Noise that disrupts SET/RESET matching has the largest impact on network performance, demonstrating how physical nonidealities translate into neural-network accuracy changes.\"}]","Integration of Physics-Derived Memristor Models with Machine Learning Frameworks - Research Report | PDF",1785901869,13,{"code":4,"msg":32,"data":33},"ok",{"site_id":25,"language":24,"slug":34,"title":13,"keywords":35,"description":14,"schema_data":36,"social_meta":88,"head_meta":90,"extra_data":92,"updated_unix":29},"integration-of-physics-derived-memristor-models-with-machine-learning-frameworks-research-report","",{"@graph":37,"@context":87},[38,55,70],{"@type":39,"itemListElement":40},"BreadcrumbList",[41,45,49,52],{"item":42,"name":43,"@type":44,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":46,"name":47,"@type":44,"position":48},"https://docshare.wps.com/document/","Document",2,{"item":50,"name":12,"@type":44,"position":51},"https://docshare.wps.com/document/research-report/",3,{"item":53,"name":13,"@type":44,"position":54},"https://docshare.wps.com/document/integration-of-physics-derived-memristor-models-with-machine-learning-frameworks-research-report/125892/",4,{"url":53,"name":13,"@type":56,"author":57,"headline":13,"publisher":59,"fileFormat":62,"inLanguage":24,"description":14,"dateModified":63,"datePublished":64,"encodingFormat":62,"isAccessibleForFree":65,"interactionStatistic":66},"DigitalDocument",{"name":9,"@type":58},"Person",{"url":42,"name":60,"@type":61},"DocShare","Organization","application/pdf","2026-08-22","2026-08-05",true,{"@type":67,"interactionType":68,"userInteractionCount":20},"InteractionCounter",{"@type":69},"ViewAction",{"@type":71,"mainEntity":72},"FAQPage",[73,79,83],{"name":74,"@type":75,"acceptedAnswer":76},"Why do existing memristor simulation frameworks struggle to model device nonidealities accurately?","Question",{"text":77,"@type":78},"They often use lookup tables or simplified analytic models, which miss physically grounded switching effects such as the cause of switching that affects switching timings and conductance-state estimation.","Answer",{"name":80,"@type":75,"acceptedAnswer":81},"What physical dynamics does the proposed work bring into the memristor model?",{"text":82,"@type":78},"It incorporates physical dynamics for VCM cells by solving an ODE that captures oxygen-vacancy changes, linking switching nonlinearity and SET/RESET asymmetry to thermal and material-related parameters.",{"name":84,"@type":75,"acceptedAnswer":85},"What is the main result from testing with the MNIST dataset?",{"text":86,"@type":78},"Noise that disrupts SET/RESET matching has the largest impact on network performance, demonstrating how physical nonidealities translate into neural-network accuracy changes.","https://schema.org",{"og:url":53,"og:type":89,"og:title":13,"og:site_name":60,"og:description":14},"article",{"robots":91,"canonical":53},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":94},[95,99,103,107,111,116,121,124,128,131,135],{"id":21,"doc_module":4,"doc_module_name":47,"category_name":96,"show_sort_weight":97,"slug":98},"Story & Novel",90,"story-novel",{"id":48,"doc_module":4,"doc_module_name":47,"category_name":100,"show_sort_weight":101,"slug":102},"Literature",80,"literature",{"id":54,"doc_module":4,"doc_module_name":47,"category_name":104,"show_sort_weight":105,"slug":106},"Exam",70,"exam",{"id":22,"doc_module":4,"doc_module_name":47,"category_name":108,"show_sort_weight":109,"slug":110},"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":47,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":47,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":47,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":20,"doc_module":4,"doc_module_name":47,"category_name":125,"show_sort_weight":126,"slug":127},"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":47,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":47,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":47,"category_name":137,"show_sort_weight":22,"slug":138},19,"General","general"]