[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85103-en":3,"doc-seo-85103-105":29,"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":13,"seo_description":14,"update_tm":27,"read_time":28},85103,687197207057,"Sage","https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0",8,"Research & Report","Steering Neural Network Training through Interpretable Constraints Based on Partial Dependence","Interest in interpretable machine learning has grown, yet explanation quality and faithfulness to prior knowledge remain underexplored, especially when adjusting models via explanation-guided learning. This work proposes steering neural network training using partial dependence so the model’s average response to selected features matches functional domain knowledge. Experiments across regression tasks, including dynamical systems forecasting, show improved performance, higher data efficiency, and explanations that align with user-provided knowledge.","arXiv :2607 .0864 1v 1 [ cs .LG] 9 Jul 2026  \nSteering Neural Network Training through Interpretable Constraints Based on Partial Dependence  \nYann Claes [y. claes@uliege. be](y. claes@uliege. be)  \nMontefiore Institute University of Liège  \nPierre Geurts  \nMontefiore Institute University of Liège  \nVân Anh Huynh-Thu  \nMontefiore Institute University of Liège  \nAbstract  \nOver the last few years, there has been an increased interest in making machine learning models more interpretable. Although a great deal of effort goes into developing techniques for interpreting the interactions learned by a given model, fewer studies focus on assessing the quality of such explanations. Even fewer focus on how to adjust the model to produce explanations faithful to prior knowledge, a process known as explanation-guided learning.  \nFurthermore, most approaches in this area focus on classification problems and usually assume prior knowledge about which input features or regions are most important. In this work, we introduce a new approach to steering neural networks based on partial dependence, such that their average response to certain features aligns with specific functional domain knowledge about the problem. We empirically demonstrate on a range of regression problems, including dynamical systems forecasting, that models whose training has been controlled using our method perform better than unconstrained models and are more dataefficient. Moreover, we highlight that interpretations obtained from the former actually align with the user-provided knowledge, whereas those obtained from the latter do not.  \n1 Introduction  \nOver the past decades, progress in machine learning (ML) methods has enabled practitioners in a variety of fields to tackle increasingly complex problems, in which such models serve as surrogates for first-principles models or costly simulators. While they come with great expressiveness, it is nevertheless not straightforward to interpret their reasoning process, which is why they are often referred to as \"black box\" models. There has been recently a vast development of the field of explainable artificial intelligence (XAI), in a desire to increase our understanding of decisions made by ML models (Adadi & Berrada, 2018; Gilpin et al., 2018; Linardatos et al., 2020) .  \nOn the one hand, models can be made more interpretable, through specific design choices. For instance, Nelder & Wedderburn (1972) introduced generalized linear models (GLM) to model various response distributions, which are more flexible than linear regression, yet still very simple to explain. Hastie & Tibshirani (1986) later implemented generalized additive models (GAM) as an extension to GLMs with non-linear predictors, and Caruana et al. (2015) successfully applied GAMs to healthcare problems.  \nOn the other hand, model predictions can be interpreted through model-specific and model-agnostic interpretation tools, with applications in different domains such as image processing (Zhang & Zhu, 2018), healthcare (Vellido, 2020) and neural language processing (Madsen et al., 2022) . For example, Lundberg  \n& Lee (2017) introduced SHAP values to compute feature importance in the prediction of the model. In a similar philosophy, Friedman (2001) designed partial dependence plots to visualize the marginal contribution of certain features to the model response, and complementary plotting functions were later implemented (Goldstein et al., 2015; Apley & Zhu, 2020) . For interpretations of visual predictions, most works rely on class activation mapping techniques to localize regions of importance (Zhou et al., 2016; Selvaraju et al. , 2017; Chattopadhay et al., 2018) .  \nAlthough current interpretation tools can help domain experts to understand how decisions are made, they cannot take their feedback into account, which could be important if the model provides a correct decision for the wrong reasons. Gao et al. (2024) notice that most works in XAI focus on the gene","cbCaioQEIdnlKCYf","https://ap.wps.com/l/cbCaioQEIdnlKCYf","pdf",2176930,1,40,"English","en",105,"# Abstract\n# Introduction\n# Related work","[{\"question\":\"What problem does the work address in explainable AI?\",\"answer\":\"It focuses on assessing explanation quality and ensuring explanations remain faithful to prior knowledge, not just generating explanations.\"},{\"question\":\"How does the proposed method steer neural network training?\",\"answer\":\"It uses partial dependence to control training so the model’s average response to certain features aligns with functional domain knowledge.\"},{\"question\":\"What evidence is provided that the method improves both models and explanations?\",\"answer\":\"Experiments on regression problems, including dynamical systems forecasting, show better performance and data efficiency than unconstrained training, and explanations that match the user-provided knowledge.\"}]",1784201124,101,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":27},"steering-neural-network-training-through-interpretable-constraints-based-on-partial-dependence","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"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":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/steering-neural-network-training-through-interpretable-constraints-based-on-partial-dependence/85103/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",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 work address in explainable AI?","Question",{"text":75,"@type":76},"It focuses on assessing explanation quality and ensuring explanations remain faithful to prior knowledge, not just generating explanations.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the proposed method steer neural network training?",{"text":80,"@type":76},"It uses partial dependence to control training so the model’s average response to certain features aligns with functional domain knowledge.",{"name":82,"@type":73,"acceptedAnswer":83},"What evidence is provided that the method improves both models and explanations?",{"text":84,"@type":76},"Experiments on regression problems, including dynamical systems forecasting, show better performance and data efficiency than unconstrained training, and explanations that match the user-provided knowledge.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":21,"slug":118},7,"Healthcare","healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]