[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-116855-en":3,"doc-seo-116855-105":30,"detail-sidebar-cat-0-en-105":83},{"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":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},116855,1099513958762,"Logic","https://ap-avatar.wpscdn.com/avatar/1000023916a998db790?x-image-process=image/resize,m_fixed,w_180,h_180&k=1784791008015729253",8,"Research & Report","PHYSICS-INSPIRED INTERPRETABILITY OF MACHINE LEARNING MODELS","Explainable AI for safety-critical domains depends on identifying which input features drive machine learning model decisions. This work introduces a physics-inspired technique that connects machine learning loss landscapes with energy landscape concepts from physical sciences. Conserved weights within groups of minima in the loss landscape are used to reveal the drivers behind model behavior. The method is motivated by analogous tools in molecular sciences such as coordinate invariants and order parameters, and is evaluated through synthetic and real-world examples.","arXiv :2304 .0238 1v 1 [ cs .LG] 5 Apr 2023  \nPHYSICS-INSPIRED INTERPRETABILITY OF MACHINE LEARNING MODELS  \nMaximilian P. Niroomand, David J. Wales  \nDepartment of Chemistry University of Cambridge fmpn26,[djw34](djw34g@cam.ac.uk)[g](djw34g@cam.ac.uk)[@cam.ac.uk](djw34g@cam.ac.uk)  \nABSTRACT  \nThe ability to explain decisions made by machine learning models remains one of the most significant hurdles towards widespread adoption of AI in highly sensitive areas such as medicine, cybersecurity or autonomous driving. Great interest exists in understanding which features of the input data prompt model decision making.  \nIn this contribution, we propose a novel approach to identify relevant features of the input data, inspired by methods from the energy landscapes field, developed in the physical sciences. By identifying conserved weights within groups of minima of the loss landscapes, we can identify the drivers of model decision making.  \nAnalogues to this idea exist in the molecular sciences, where coordinate invariants or order parameters are employed to identify critical features of a molecule.  \nHowever, no such approach exists for machine learning loss landscapes. We will demonstrate the applicability of energy landscape methods to machine learning models and give examples, both synthetic and from the real world, for how these methods can help to make models more interpretable.  \n1 INTRODUCTION  \nMachine learning methods have achieved impressive results in recent years. Besides famous applications in areas like chess (Silver et al., 2017a) and Go (Silver et al., 2017b), AI plays a critical role in advances to autonomous driving (Grigorescu et al., 2020), protein structure prediction (Jumper et al., 2021), cancer identification (Sammut et al., 2022) and in cybersecurity (Dasgupta et al., 2022) . However, in order for AI methods to take the next step and be commonly employed for critical applications without any humans in the loop, we want to be able to understand the decision making process. A critical component towards explainable AI is understanding which parts of the input data are utilised by the model in its decision making. In neural networks, the most popular approach is to study the outgoing weights and gradients from an individual input node. Larger weights are reasonably assumed to indicate a greater significance of the particular input, and indeed, an entire class of interpretability metrics, namely gradient-based methods, are founded on this idea (Simonyan et al., 2013; Linardatos et al., 2020) . Yet, given the immense complexity of overparameterised, deep neural networks, current methods are in practice often insufficient to appropriately explain a model. Using methods from the physical sciences, we propose a novel approach as a next step towards interpretable neural networks.  \n1.1 ENERGY LANDSCAPES  \nIn the physical sciences, energy landscapes (ELs) are employed to explore molecular configuration space (Wales et al., 1998; 2003) . Each molecular configuration is associated with an energy value, and local minima of the energy landscape represent stable isomers. The analogy to machine learning loss landscapes (ML-LLs) is straightforward, the main difference perhaps being that non-minima are valid configurations for sets of weights. Due to this similarity between ELs and ML-LLs, various, well-established methods from the field of energy landscapes can be employed to study ML-LLs. One key area of interest here is interpretability. Employing well-understood methods from a mature field, with a solid mathematical basis in the physical world, to move away from black-box machine learning models may be a helpful step towards interpretable machine learning models.  \n1.2 RELATED WORK  \nVarious approaches to interpretability in deep learning for neural networks exist. Below, we are mostly interested in gradient-based methods due to their applicability to non-image data. Various other methods to interpret the output of CNNs on images ","cbCaiuSrPeNiBLAj","https://ap.wps.com/l/cbCaiuSrPeNiBLAj","pdf",261368,1,6,"English","en",105,"# Abstract\n# Introduction\n## Energy Landscapes\n## Related Work\n## Gradient-based Methods\n## Energy Landscapes in Machine Learning\n## Interpreting Energy Landscapes","[{\"question\":\"What evidence is provided to support the method?\",\"answer\":\"The paper demonstrates applicability using examples from both synthetic data and real-world settings, showing how energy landscape methods can improve interpretability.\"}]","PHYSICS-INSPIRED INTERPRETABILITY OF MACHINE LEARNING MODELS | PDF",1785672092,15,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":78,"head_meta":80,"extra_data":82,"updated_unix":28},"physics-inspired-interpretability-of-machine-learning-models","",{"@graph":36,"@context":77},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/physics-inspired-interpretability-of-machine-learning-models/116855/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-02",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71],{"name":72,"@type":73,"acceptedAnswer":74},"What evidence is provided to support the method?","Question",{"text":75,"@type":76},"The paper demonstrates applicability using examples from both synthetic data and real-world settings, showing how energy landscape methods can improve interpretability.","Answer","https://schema.org",{"og:url":52,"og:type":79,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":81,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":84},[85,89,93,97,102,106,111,114,119,122,126],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":86,"show_sort_weight":87,"slug":88},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":90,"show_sort_weight":91,"slug":92},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Exam",70,"exam",{"id":98,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},5,"Comic",60,"comic",{"id":21,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Technology",50,"technology",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":112,"slug":113},30,"research-report",{"id":115,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},9,"Religion & Spirituality",20,"religion-spirituality",{"id":117,"doc_module":4,"doc_module_name":46,"category_name":120,"show_sort_weight":117,"slug":121},"World Cup","world-cup",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":123,"slug":125},10,"Lifestyle","lifestyle",{"id":127,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":98,"slug":129},19,"General","general"]