[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-122062-en":3,"doc-seo-122062-105":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":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},122062,8796095461564,"Liam","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","Integrating White and Black Box Techniques for Interpretable Machine Learning","Machine learning faces an accuracy–interpretability trade-off: models that are easy for humans to understand often achieve lower performance than more complex, opaque methods. This paper proposes an ensemble classifier that routes “easy” inputs to a highly interpretable white-box model and “hard” inputs to a stronger black-box model. A grader classifier labels inputs as easy or hard, enabling understandable explanations or justified deferrals, while training base and deferral models separately and learning grader labels from their outcomes.","arXiv :2407 .08973v 1 [ cs .LG] 12 Jul 2024  \nIntegrating White and Black Box Techniques for Interpretable Machine Learning  \nEric M. Vernon, Naoki Masuyama, and Yusuke Nojima  \nOsaka Metropolitan University, Sakai, Osaka 5998531, Japan,{sn22864k@st., masuyama@, [nojima@](nojima@}omu.ac.jp)[}](nojima@}omu.ac.jp)[omu.ac.jp](nojima@}omu.ac.jp)  \nAbstract. In machine learning algorithm design, there exists a tradeoff between the interpretability and performance of the algorithm. In general, algorithms which are simpler and easier for humans to comprehend tend to show worse performance than more complex, less transparent algorithms. For example, a random forest classifier is likely to be more accurate than a simple decision tree, but at the expense of interpretability. In this paper, we present an ensemble classifier design which classifies easier inputs using a highly-interpretable classifier (i.e., white box model), and more difficult inputs using a more powerful, but less interpretable classifier (i.e., black box model) .  \nKeywords: machine learning, classification, explainable artificial intelligence, accuracy-interpretability trade-off  \n1 Introduction  \nOne of the most pressing issues in machine learning (ML) research today is the concern that many popular ML algorithms operate as a “black box” -that is, they offer no human-understandable explanation for their outputs. Generally speaking, there is a trade-off (the so-called “accuracy-interpretability trade-off”) between the performance of an ML model and how easily a human can understand the steps taken to reach a given conclusion.  \nFor example, this trade-off can be seen quite plainly in decision trees [1] . A shallow decision tree with only a few branches is quite easy to understand, but is more limited in its ability to describe complex datasets. Increasing the depth of the tree will generally improve accuracy, but at the cost of interpretability. Ensemble methods such as random forests [2] or gradient boosted trees [3] are generally even more accurate while further obscuring the decision making.  \nIn this paper, we present an ensemble classifier design which uses a simple, easily understood classifier to classify “easy” inputs and a more complex classifier for “hard” inputs. The final piece of the ensemble is a “grader” classifier which classifies inputs as either “easy” or “hard”.  \nTo classify a new input using our design, it is first evaluated by the grader. If the output is “easy”, then the pattern is evaluated by the “base classifier” (e.g. , a decision tree) . If the output is “hard”, then the pattern is evaluated by the“deferral classifier”(e.g., a random forest) . In our experiments, we use a decision  \n2 E. M. Vernon et al.  \ntree classifier for the grader as well. This means that the user will either be able to directly understand the reasoning behind system output (in the case of easy inputs), or be given an understandable reason for why a more complex classifier was required (in the case of hard inputs) .  \nTo train the ensemble, the base and deferral classifiers are independently fit to the training data, as per normal. Then, the training data copied and assigned new labels, either “easy” or “hard”: Easy inputs are patterns which are correctly classified by the base classifier, hard inputs are those which are not. The grader is then fit to this modified training set.  \nThis paper is organized as follows: In Section 2, we describe the background of our research. Section 3 gives a detailed overview of our proposed method, including a worked example using a synthetic 2-D dataset. Section 4 describes the computational experiments performed to validate our approach and the associated results; Section 5 concludes the paper.  \n2 Research Background  \n2.1 Interpretability in Machine Learning  \nMachine learning algorithms are everywhere: They personalize the content we see on the web, secure our bank accounts against fraud, and assist our doctors in diagnosing our ","cbCaiqgf1z332j8C","https://ap.wps.com/l/cbCaiqgf1z332j8C","pdf",558712,1,10,"English","en",105,"# Introduction\n## Interpretability and the accuracy–interpretability trade-off\n## Classification with a reject option\n# Research Background\n## Interpretability in Machine Learning\n## Classification with a Reject Option\n# Proposed Method\n## Ensemble design with easy/hard routing\n## Grader, base classifier, and deferral classifier\n# Experiments and Results\n## Computational experiments\n# Conclusion","[{\"question\":\"What problem does the paper address in interpretable machine learning?\",\"answer\":\"It addresses the accuracy–interpretability trade-off, where simpler models that humans can understand may perform worse than complex black-box models.\"},{\"question\":\"How does the proposed ensemble handle easy versus hard inputs?\",\"answer\":\"A grader classifier first labels an input as easy or hard; easy inputs are processed by a highly interpretable base classifier, while hard inputs are processed by a more powerful but less interpretable deferral classifier.\"},{\"question\":\"How is the ensemble trained to create the grader’s easy/hard labels?\",\"answer\":\"The base and deferral classifiers are trained normally, then training data are relabeled as easy if the base classifier correctly classifies them and hard otherwise; the grader is then fit on this modified dataset.\"}]","Integrating White and Black Box Techniques for Interpretable Machine Learning | PDF",1785808622,25,{"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},"integrating-white-and-black-box-techniques-for-interpretable-machine-learning","",{"@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/integrating-white-and-black-box-techniques-for-interpretable-machine-learning/122062/",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","2026-08-04",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the paper address in interpretable machine learning?","Question",{"text":75,"@type":76},"It addresses the accuracy–interpretability trade-off, where simpler models that humans can understand may perform worse than complex black-box models.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the proposed ensemble handle easy versus hard inputs?",{"text":80,"@type":76},"A grader classifier first labels an input as easy or hard; easy inputs are processed by a highly interpretable base classifier, while hard inputs are processed by a more powerful but less interpretable deferral classifier.",{"name":82,"@type":73,"acceptedAnswer":83},"How is the ensemble trained to create the grader’s easy/hard labels?",{"text":84,"@type":76},"The base and deferral classifiers are trained normally, then training data are relabeled as easy if the base classifier correctly classifies them and hard otherwise; the grader is then fit on this modified dataset.","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},[93,97,101,105,110,115,120,123,128,131,134],{"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":21,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":21,"slug":133},"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]