[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119245-en":3,"doc-seo-119245-105":29,"detail-sidebar-cat-0-en-105":89},{"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":11},119245,687197207639,"Asher","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","Interpretable Machine Learning via Linear Temporal Logic","Deep neural networks deliver strong predictive accuracy but remain difficult to interpret, limiting adoption in high-stakes domains such as medicine and safety-critical control. This work introduces a learning framework that represents decisions as inherently interpretable machine learning models by learning Linear Temporal Logic (LTL) formulas. The framework supports both supervised learning for time-series classification and unsupervised learning for anomaly detection, while enabling extensions for noisy data and incorporation of expert knowledge. ","Interpretable Machine Learning via Linear Temporal Logic  \nSimon Lutz [simon.lutz@tu-dortmund.de](simon.lutz@tu-dortmund.de)  \nResearch Center \\Trustworthy Data Science and Security\", University Alliance Ruhr, TU Dortmund University, Germany  \nDaniel Neider daniel.neider@tu-dortmund.de  \nProfessor for Veri􀀌cation and Formal Guarantees of Machine Learning, Research Center \\Trustworthy Data Science and Security\", University Alliance Ruhr, TU Dortmund University, Germany  \nAbstract  \nIn recent years, deep neural networks have shown excellent performance, outperforming even human experts in various tasks. However, their inherent complexity and black-box nature often make it hard, if not impossible, to understand the decisions made by these models, hindering their practical application in high-stakes scenarios. We propose a framework for learning LTL formulas as inherently interpretable machine learning models. These models can be trained both in a supervised and unsupervised setting. Furthermore, they can easily be extended to handle noisy data and to incorporate expert knowledge.  \nKeywords: Explainable AI, Learning of logic formulas, Linear Temporal Logic  \nIn the last years, Arti􀀌cial Intelligence (AI) has received tremendous attention and is nowadays used in a wide variety of application domains including medicine, law enforcement, autonomous systems, and natural language processing, to name but a few. Inmost cases, these AI systems are based on deep neural networks with hundreds of layers and billions of parameters. Trained on large amounts of training data, these models have shown excellent performance, outperforming even human experts in various tasks. However, their inherent complexity and black-box nature often make it hard, if not impossible, to understand the decisions made by a neural network. This is especially problematic in high-stakes application and often a severe obstacle to employing AI systems in practice. Consider, for instance, a medical system assisting doctors with diagnosing patients. If the system diagnoses a speci􀀌c disease, it is imperative to understand the reason to ensure correct treatment is prescribed.  \nTo overcome this drawback of intransparent decision-making, the 􀀌eld of explainable arti􀀌cial in-  \ntelligence (XAI) has evolved in recent years. Instead of just computing the decision of a neural network, XAI methods also provide a human-readable explanation of how the network concluded this decision (see [1] for a more detailed introduction) . Broadly speaking, these methods can be separated into post hoc explanations and inherently interpretable models. Post hoc explanations do not interfere with the architecture or training of a neural network but aim at inferring an explanation by analyzing its decisionmaking post hoc. State-of-the-art methods include using game-theoretic Shapley values as a measure for feature importance (SHAP [2]), the use of surrogate models for local explanations (LIME [3]), and the visualization of feature importance using heat maps (Grad-CAM [4]), to name but a few. Instead of training a complex neural network, the second paradigm opts for training simpler models such as decision trees, decision rules, or linear regression. Even though these models are inherently interpretable, they may lack the ability to generalize well from the training data leading to worse overall performance. Nevertheless, multiple papers have recently introduced deterministic 􀀌nite automata (DFAs) as capable (i.e., on par with state-of-the-art LSTM models) yet interpretable models for sequence classi􀀌cation [5, 6] and anomaly detection [7] .  \nIn this paper, we follow the second paradigm and introduce a framework for learning formulas in Linear Temporal Logic (LTL) [8] as interpretable machine learning models for time series data. This speci􀀌c choice of model is motivated by the following observations: 􀀌rst, a description of the observed, temporal behavior can often be captured in a concise logica","cbCaina4tqm36MuU","https://ap.wps.com/l/cbCaina4tqm36MuU","pdf",175770,1,3,"English","en",105,"# Introduction\n## Limitations of black-box neural networks\n## Explainable AI: post hoc vs inherently interpretable models\n## Linear Temporal Logic as an interpretable model\n# Learning Framework\n## Supervised setup: LTL-based classification\n## One-vs-rest minimal LTL formulas\n## Bayesian inference for posterior class probabilities\n# Unsupervised Setup\n## Anomaly detection with LTL formulas\n## Extensions for noisy data and expert knowledge","[{\"question\":\"Why are black-box neural networks hard to use in high-stakes scenarios?\",\"answer\":\"Their complexity makes it difficult to understand why a specific decision was made, which can block reliable use in domains like medicine and safety-critical applications.\"},{\"question\":\"What does the proposed framework learn?\",\"answer\":\"It learns Linear Temporal Logic (LTL) formulas that act as inherently interpretable machine learning models for time series.\"},{\"question\":\"How does the framework perform supervised learning for classification?\",\"answer\":\"It builds minimal LTL formulas in a one-vs-rest manner for each class, then queries them to obtain decision distributions and uses an approximate Bayesian method to infer posterior class probabilities.\"}]","Interpretable Machine Learning via Linear Temporal Logic | PDF",1785723274,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":84,"head_meta":86,"extra_data":88,"updated_unix":28},"interpretable-machine-learning-via-linear-temporal-logic","",{"@graph":35,"@context":83},[36,52,66],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,49],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":21},"https://docshare.wps.com/document/research-report/",{"item":50,"name":13,"@type":42,"position":51},"https://docshare.wps.com/document/interpretable-machine-learning-via-linear-temporal-logic/119245/",4,{"url":50,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"inLanguage":23,"description":14,"dateModified":60,"datePublished":60,"encodingFormat":59,"isAccessibleForFree":61,"interactionStatistic":62},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":40,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":63,"interactionType":64,"userInteractionCount":4},"InteractionCounter",{"@type":65},"ViewAction",{"@type":67,"mainEntity":68},"FAQPage",[69,75,79],{"name":70,"@type":71,"acceptedAnswer":72},"Why are black-box neural networks hard to use in high-stakes scenarios?","Question",{"text":73,"@type":74},"Their complexity makes it difficult to understand why a specific decision was made, which can block reliable use in domains like medicine and safety-critical applications.","Answer",{"name":76,"@type":71,"acceptedAnswer":77},"What does the proposed framework learn?",{"text":78,"@type":74},"It learns Linear Temporal Logic (LTL) formulas that act as inherently interpretable machine learning models for time series.",{"name":80,"@type":71,"acceptedAnswer":81},"How does the framework perform supervised learning for classification?",{"text":82,"@type":74},"It builds minimal LTL formulas in a one-vs-rest manner for each class, then queries them to obtain decision distributions and uses an approximate Bayesian method to infer posterior class probabilities.","https://schema.org",{"og:url":50,"og:type":85,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":87,"canonical":50},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":90},[91,95,99,103,108,113,118,121,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":92,"show_sort_weight":93,"slug":94},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":96,"show_sort_weight":97,"slug":98},"Literature",80,"literature",{"id":51,"doc_module":4,"doc_module_name":45,"category_name":100,"show_sort_weight":101,"slug":102},"Exam",70,"exam",{"id":104,"doc_module":4,"doc_module_name":45,"category_name":105,"show_sort_weight":106,"slug":107},5,"Comic",60,"comic",{"id":109,"doc_module":4,"doc_module_name":45,"category_name":110,"show_sort_weight":111,"slug":112},6,"Technology",50,"technology",{"id":114,"doc_module":4,"doc_module_name":45,"category_name":115,"show_sort_weight":116,"slug":117},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":119,"slug":120},30,"research-report",{"id":122,"doc_module":4,"doc_module_name":45,"category_name":123,"show_sort_weight":124,"slug":125},9,"Religion & Spirituality",20,"religion-spirituality",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":124,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":104,"slug":136},19,"General","general"]