[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-124937-en":3,"doc-seo-124937-105":30,"detail-sidebar-cat-0-en-105":91},{"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":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},124937,5909877438554,"Maeve","https://ap-avatar.wpscdn.com/avatar/5600025385ad2bf12a7?_k=1778553567797529272",8,"Research & Report","On the Predictive Accuracy of Neural Temporal Point Process Models for Continuous-time Event Data - Comprehensive large-scale study","Temporal Point Processes (TPPs) provide a core mathematical framework for modeling asynchronous event sequences in continuous time, but classical approaches rely on restrictive assumptions that limit their ability to represent complex dynamics. Neural TPPs introduce neural-network parameterizations to improve flexibility and efficiency, yet existing work often uses inconsistent baselines, datasets, and experimental setups, obscuring the drivers of predictive accuracy. This study conducts a comprehensive unified large-scale evaluation across real and synthetic datasets, analyzing event encoding, history encoders, decoder parametrizations, and probabilistic calibration, with findings on history size effects and mark distribution miscalibration.","arXiv :2306 . 17066v1 [ cs .LG] 29 Jun 2023  \nOn the Predictive Accuracy of Neural Temporal Point Process Models for Continuous-time Event Data  \nTanguy Bosser [tanguy. bosser@umons. ac. be](tanguy. bosser@umons. ac. be)  \nDepartment of Computer Science University of Mons  \nSouhaib Ben Taieb [souhaib. bentaieb@umons. ac. be](souhaib. bentaieb@umons. ac. be)  \nDepartment of Computer Science University of Mons  \nAbstract  \nTemporal Point Processes (TPPs) serve as the standard mathematical framework for modeling asynchronous event sequences in continuous time. However, classical TPP models are often constrained by strong assumptions, limiting their ability to capture complex real-world event dynamics. To overcome this limitation, researchers have proposed Neural TPPs, which leverage neural network parametrizations to offer more flexible and efficient modeling. While recent studies demonstrate the effectiveness of Neural TPPs, they often lack a unified setup, relying on different baselines, datasets, and experimental configurations. This makes it challenging to identify the key factors driving improvements in predictive accuracy, hindering research progress. To bridge this gap, we present a comprehensive large-scale experimental study that systematically evaluates the predictive accuracy of stateof-the-art neural TPP models. Our study encompasses multiple real-world and synthetic event sequence datasets, following a carefully designed unified setup. We thoroughly investigate the influence of major architectural components such as event encoding, history encoder, and decoder parametrization on both time and mark prediction tasks. Additionally, we delve into the less explored area of probabilistic calibration for neural TPP models. By analyzing our results, we draw insightful conclusions regarding the significance of history size and the impact of architectural components on predictive accuracy. Furthermore, we shed light on the miscalibration of mark distributions in neural TPP models. Our study aims to provide valuable insights into the performance and characteristics of neural TPP models, contributing to a better understanding of their strengths and limitations.  \n1 Introduction  \nFrom human social activity to natural phenomena, the evolution of a system of interest can often be characterized by a sequence of discrete events occurring at irregular time intervals. Online shopping activity (Cai et al., 2018), earthquake occurrences (Ogata, 1998), measurement of electronic health records (Wang et al., 2016), and users activity on social media (Farajtabar et al., 2015) are  \ntypical examples where such sequences are frequently encountered. Given a sequence of observed historical events, a crucial challenge in numerous applications is to predict the timing of future events. Additionally, in cases where events are assigned a label, referred to as marks, it is necessary to also predict the type of event that is likely to occur. It is reasonable to assume that events are interdependent and that the future evolution of a system is directly influenced by past occurrences. For example, an individual might be inclined to purchase a particular item at a specific time on ane-commerce platform solely because a previous purchase made it necessary. Hence, modeling the intricate dynamics of event occurrences becomes crucial in predicting future events based on past observations.  \nDrawing upon solid theoretical foundations, the framework of Temporal Point Processes (TPPs)(Daley & Vere-Jones, 2007) has established itself as a suitable choice for modeling these sequences of asynchronous and time-dependent data. A TPP is fully characterized by its conditional intensity function, which provides the instantaneous unit rate of event arrivals based on the process history (Rasmussen, 2018) . For event sequence modeling in a variety of domains, including finance (Hawkes, 2018), crime (Egesdal et al., 2010), or epidemiology (Rizoiu et al., 2018), a large nu","cbCaicqvHC7PiTu3","https://ap.wps.com/l/cbCaicqvHC7PiTu3","pdf",3725663,1,58,"English","en",105,"# Introduction\n## Predicting future event time and marks\n## Temporal Point Processes and neural TPPs\n## Architectural components in neural TPP models\n## Motivation for unified large-scale evaluation","[{\"question\":\"What problem does the study address in neural temporal point process research?\",\"answer\":\"Existing Neural TPP studies often use different baselines, datasets, and experimental configurations, making it difficult to determine which factors actually improve predictive accuracy.\"},{\"question\":\"Which architectural components are systematically evaluated?\",\"answer\":\"The study investigates event encoding, history encoder design, and decoder parametrization, and how these choices affect time and mark prediction.\"},{\"question\":\"What additional aspect beyond prediction accuracy is analyzed?\",\"answer\":\"The research examines probabilistic calibration, including the miscalibration of mark distributions in neural TPP models.\"}]","On the Predictive Accuracy of Neural Temporal Point Process Models for Continuous-time Event Data - Comprehensive large-scale study | PDF",1785895489,146,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":28},"on-the-predictive-accuracy-of-neural-temporal-point-process-models-for-continuous-time-event-data-comprehensive-large-scale-study","",{"@graph":36,"@context":85},[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/on-the-predictive-accuracy-of-neural-temporal-point-process-models-for-continuous-time-event-data-comprehensive-large-scale-study/124937/",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","2026-08-05",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the study address in neural temporal point process research?","Question",{"text":75,"@type":76},"Existing Neural TPP studies often use different baselines, datasets, and experimental configurations, making it difficult to determine which factors actually improve predictive accuracy.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"Which architectural components are systematically evaluated?",{"text":80,"@type":76},"The study investigates event encoding, history encoder design, and decoder parametrization, and how these choices affect time and mark prediction.",{"name":82,"@type":73,"acceptedAnswer":83},"What additional aspect beyond prediction accuracy is analyzed?",{"text":84,"@type":76},"The research examines probabilistic calibration, including the miscalibration of mark distributions in neural TPP models.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]