Artificial intelligence (AI) has seen a steady increase in use in the health and medical field, where it is used by lay users and health experts alike. However, these AI systems often lack transparency regarding the inputs and decision making process (often called black boxes), which in turn can be detrimental to the user’s satisfaction and trust towards these systems. Explainable AI (XAI) aims to overcome this problem by opening up certain aspects of the black box, and has proven to be a successful means of increasing trust, transparency and even system effectiveness. However, for certain groups (i.e. lay users in health), explanation methods and evaluation metrics still remain underexplored. In this paper, we will outline our research regarding designing and evaluating explanations for health recommendations for lay users and domain experts, as well as list a few takeaways we were already able to find in our initial studies.
{"title":"Designing and evaluating explainable AI for non-AI experts: challenges and opportunities","authors":"Maxwell Szymanski, K. Verbert, V. Abeele","doi":"10.1145/3523227.3547427","DOIUrl":"https://doi.org/10.1145/3523227.3547427","url":null,"abstract":"Artificial intelligence (AI) has seen a steady increase in use in the health and medical field, where it is used by lay users and health experts alike. However, these AI systems often lack transparency regarding the inputs and decision making process (often called black boxes), which in turn can be detrimental to the user’s satisfaction and trust towards these systems. Explainable AI (XAI) aims to overcome this problem by opening up certain aspects of the black box, and has proven to be a successful means of increasing trust, transparency and even system effectiveness. However, for certain groups (i.e. lay users in health), explanation methods and evaluation metrics still remain underexplored. In this paper, we will outline our research regarding designing and evaluating explanations for health recommendations for lay users and domain experts, as well as list a few takeaways we were already able to find in our initial studies.","PeriodicalId":443279,"journal":{"name":"Proceedings of the 16th ACM Conference on Recommender Systems","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128900489","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yingcan Wei, Matthias Langer, F. Yu, Minseok Lee, Jie Liu, Ji Shi, Zehuan Wang
Recommendation systems are of crucial importance for a variety of modern apps and web services, such as news feeds, social networks, e-commerce, search, etc. To achieve peak prediction accuracy, modern recommendation models combine deep learning with terabyte-scale embedding tables to obtain a fine-grained representation of the underlying data. Traditional inference serving architectures require deploying the whole model to standalone servers, which is infeasible at such massive scale. In this paper, we provide insights into the intriguing and challenging inference domain of online recommendation systems. We propose the HugeCTR Hierarchical Parameter Server (HPS), an industry-leading distributed recommendation inference framework, that combines a high-performance GPU embedding cache with an hierarchical storage architecture, to realize low-latency retrieval of embeddings for online model inference tasks. Among other things, HPS features (1) a redundant hierarchical storage system, (2) a novel high-bandwidth cache to accelerate parallel embedding lookup on NVIDIA GPUs, (3) online training support and (4) light-weight APIs for easy integration into existing large-scale recommendation workflows. To demonstrate its capabilities, we conduct extensive studies using both synthetically engineered and public datasets. We show that our HPS can dramatically reduce end-to-end inference latency, achieving 5~62x speedup (depending on the batch size) over CPU baseline implementations for popular recommendation models. Through multi-GPU concurrent deployment, the HPS can also greatly increase the inference QPS.
{"title":"A GPU-specialized Inference Parameter Server for Large-Scale Deep Recommendation Models","authors":"Yingcan Wei, Matthias Langer, F. Yu, Minseok Lee, Jie Liu, Ji Shi, Zehuan Wang","doi":"10.1145/3523227.3546765","DOIUrl":"https://doi.org/10.1145/3523227.3546765","url":null,"abstract":"Recommendation systems are of crucial importance for a variety of modern apps and web services, such as news feeds, social networks, e-commerce, search, etc. To achieve peak prediction accuracy, modern recommendation models combine deep learning with terabyte-scale embedding tables to obtain a fine-grained representation of the underlying data. Traditional inference serving architectures require deploying the whole model to standalone servers, which is infeasible at such massive scale. In this paper, we provide insights into the intriguing and challenging inference domain of online recommendation systems. We propose the HugeCTR Hierarchical Parameter Server (HPS), an industry-leading distributed recommendation inference framework, that combines a high-performance GPU embedding cache with an hierarchical storage architecture, to realize low-latency retrieval of embeddings for online model inference tasks. Among other things, HPS features (1) a redundant hierarchical storage system, (2) a novel high-bandwidth cache to accelerate parallel embedding lookup on NVIDIA GPUs, (3) online training support and (4) light-weight APIs for easy integration into existing large-scale recommendation workflows. To demonstrate its capabilities, we conduct extensive studies using both synthetically engineered and public datasets. We show that our HPS can dramatically reduce end-to-end inference latency, achieving 5~62x speedup (depending on the batch size) over CPU baseline implementations for popular recommendation models. Through multi-GPU concurrent deployment, the HPS can also greatly increase the inference QPS.","PeriodicalId":443279,"journal":{"name":"Proceedings of the 16th ACM Conference on Recommender Systems","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128241467","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Recommendation system often suffers from popularity bias. Often the training data inherently exhibits long-tail distribution in item popularity (data bias). Moreover, the recommendation systems could give unfairly higher recommendation scores to popular items even among items a user equally liked, resulting in over-recommendation of popular items (model bias). In this study we propose a novel method to reduce the model bias while maintaining accuracy by directly regularizing the recommendation scores to be equal across items a user preferred. Akin to contrastive learning, we extend the widely used pairwise loss (BPR loss) which maximizes the score differences between preferred and unpreferred items, with a regularization term that minimizes the score differences within preferred and unpreferred items, respectively, thereby achieving both high debias and high accuracy performance with no additional training. To test the effectiveness of the proposed method, we design an experiment using a synthetic dataset which induces model bias with baseline training; we showed applying the proposed method resulted in drastic reduction of model bias while maintaining accuracy. Comprehensive comparison with earlier debias methods showed the proposed method had advantages in terms of computational validity and efficiency. Further empirical experiments utilizing four benchmark datasets and four recommendation models indicated the proposed method showed general improvements over performances of earlier debias methods. We hope that our method could help users enjoy diverse recommendations promoting serendipitous findings. Code available at https://github.com/stillpsy/popbias.
{"title":"Countering Popularity Bias by Regularizing Score Differences","authors":"Wondo Rhee, S. Cho, B. Suh","doi":"10.1145/3523227.3546757","DOIUrl":"https://doi.org/10.1145/3523227.3546757","url":null,"abstract":"Recommendation system often suffers from popularity bias. Often the training data inherently exhibits long-tail distribution in item popularity (data bias). Moreover, the recommendation systems could give unfairly higher recommendation scores to popular items even among items a user equally liked, resulting in over-recommendation of popular items (model bias). In this study we propose a novel method to reduce the model bias while maintaining accuracy by directly regularizing the recommendation scores to be equal across items a user preferred. Akin to contrastive learning, we extend the widely used pairwise loss (BPR loss) which maximizes the score differences between preferred and unpreferred items, with a regularization term that minimizes the score differences within preferred and unpreferred items, respectively, thereby achieving both high debias and high accuracy performance with no additional training. To test the effectiveness of the proposed method, we design an experiment using a synthetic dataset which induces model bias with baseline training; we showed applying the proposed method resulted in drastic reduction of model bias while maintaining accuracy. Comprehensive comparison with earlier debias methods showed the proposed method had advantages in terms of computational validity and efficiency. Further empirical experiments utilizing four benchmark datasets and four recommendation models indicated the proposed method showed general improvements over performances of earlier debias methods. We hope that our method could help users enjoy diverse recommendations promoting serendipitous findings. Code available at https://github.com/stillpsy/popbias.","PeriodicalId":443279,"journal":{"name":"Proceedings of the 16th ACM Conference on Recommender Systems","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127925969","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Re-ranking is one of the most critical stages for multi-stage recommender systems (MRS), which re-orders the input ranking lists by modeling the cross-item interaction. Recent re-ranking methods have evolved into deep neural architectures due to the significant advances in deep learning. Neural re-ranking, therefore, has become a trending topic and many of the improved algorithms have demonstrated their use in industrial applications, enjoying great commercial success. The purpose of this tutorial is to explore some of the recent work on neural re-ranking, integrating them into a broader picture and paving ways for more comprehensive solutions for future research. In particular, we provide a taxonomy of current methods according to the objectives and training signals. We examine and compare these methods qualitatively and quantitatively, and identify some open challenges and future prospects.
{"title":"Neural Re-ranking for Multi-stage Recommender Systems","authors":"Weiwen Liu, Jiarui Qin, Ruiming Tang, Bo Chen","doi":"10.1145/3523227.3547369","DOIUrl":"https://doi.org/10.1145/3523227.3547369","url":null,"abstract":"Re-ranking is one of the most critical stages for multi-stage recommender systems (MRS), which re-orders the input ranking lists by modeling the cross-item interaction. Recent re-ranking methods have evolved into deep neural architectures due to the significant advances in deep learning. Neural re-ranking, therefore, has become a trending topic and many of the improved algorithms have demonstrated their use in industrial applications, enjoying great commercial success. The purpose of this tutorial is to explore some of the recent work on neural re-ranking, integrating them into a broader picture and paving ways for more comprehensive solutions for future research. In particular, we provide a taxonomy of current methods according to the objectives and training signals. We examine and compare these methods qualitatively and quantitatively, and identify some open challenges and future prospects.","PeriodicalId":443279,"journal":{"name":"Proceedings of the 16th ACM Conference on Recommender Systems","volume":"67 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125491344","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Dmytro Ivchenko, D. V. Staay, Colin Taylor, Xing Liu, Will Feng, Rahul Kindi, Anirudh Sudarshan, S. Sefati
Recommendation Systems (RecSys) comprise a large footprint of production-deployed AI today. The neural network-based recommender systems differ from deep learning models in other domains in using high-cardinality categorical sparse features that require large embedding tables to be trained. In this talk we introduce TorchRec, a PyTorch domain library for Recommendation Systems. This new library provides common sparsity and parallelism primitives, enabling researchers to build state-of-the-art personalization models and deploy them in production. In this talk we cover the building blocks of the TorchRec library including modeling primitives such as embedding bags and jagged tensors, optimized recommender system kernels powered by FBGEMM, a flexible sharder that supports a veriety of strategies for partitioning embedding tables, a planner that automatically generates optimized and performant sharding plans, support for GPU inference and common modeling modules for building recommender system models. TorchRec library is currently used to train large-scale recommender models at Meta. We will present how TorchRec helped Meta’s recommender system platform to transition from CPU asynchronous training to accelerator-based full-sync training.
{"title":"TorchRec: a PyTorch Domain Library for Recommendation Systems","authors":"Dmytro Ivchenko, D. V. Staay, Colin Taylor, Xing Liu, Will Feng, Rahul Kindi, Anirudh Sudarshan, S. Sefati","doi":"10.1145/3523227.3547387","DOIUrl":"https://doi.org/10.1145/3523227.3547387","url":null,"abstract":"Recommendation Systems (RecSys) comprise a large footprint of production-deployed AI today. The neural network-based recommender systems differ from deep learning models in other domains in using high-cardinality categorical sparse features that require large embedding tables to be trained. In this talk we introduce TorchRec, a PyTorch domain library for Recommendation Systems. This new library provides common sparsity and parallelism primitives, enabling researchers to build state-of-the-art personalization models and deploy them in production. In this talk we cover the building blocks of the TorchRec library including modeling primitives such as embedding bags and jagged tensors, optimized recommender system kernels powered by FBGEMM, a flexible sharder that supports a veriety of strategies for partitioning embedding tables, a planner that automatically generates optimized and performant sharding plans, support for GPU inference and common modeling modules for building recommender system models. TorchRec library is currently used to train large-scale recommender models at Meta. We will present how TorchRec helped Meta’s recommender system platform to transition from CPU asynchronous training to accelerator-based full-sync training.","PeriodicalId":443279,"journal":{"name":"Proceedings of the 16th ACM Conference on Recommender Systems","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134131939","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
V. W. Anelli, Pierpaolo Basile, Gerard de Melo, F. Donini, Antonio Ferrara, C. Musto, F. Narducci, A. Ragone, M. Zanker
In the last few years, a renewed interest of the research community in conversational recommender systems (CRSs) has been emerging. This is likely due to the massive proliferation of Digital Assistants (DAs) such as Amazon Alexa, Siri, or Google Assistant that are revolutionizing the way users interact with machines. DAs allow users to execute a wide range of actions through an interaction mostly based on natural language utterances. However, although DAs are able to complete tasks such as sending texts, making phone calls, or playing songs, they still remain at an early stage in terms of their recommendation capabilities via a conversation. In addition, we have been witnessing the advent of increasingly precise and powerful recommendation algorithms and techniques able to effectively assess users’ tastes and predict information that may be of interest to them. Most of these approaches rely on the collaborative paradigm (often exploiting machine learning techniques) and neglect the huge amount of knowledge, both structured and unstructured, describing the domain of interest of a recommendation engine. Although very effective in predicting relevant items, collaborative approaches miss some very interesting features that go beyond the accuracy of results and move in the direction of providing novel and diverse results as well as generating explanations for recommended items. Knowledge-aware side information becomes crucial when a conversational interaction is implemented, in particular for preference elicitation, explanation, and critiquing steps.
{"title":"Fourth Knowledge-aware and Conversational Recommender Systems Workshop (KaRS)","authors":"V. W. Anelli, Pierpaolo Basile, Gerard de Melo, F. Donini, Antonio Ferrara, C. Musto, F. Narducci, A. Ragone, M. Zanker","doi":"10.1145/3523227.3547412","DOIUrl":"https://doi.org/10.1145/3523227.3547412","url":null,"abstract":"In the last few years, a renewed interest of the research community in conversational recommender systems (CRSs) has been emerging. This is likely due to the massive proliferation of Digital Assistants (DAs) such as Amazon Alexa, Siri, or Google Assistant that are revolutionizing the way users interact with machines. DAs allow users to execute a wide range of actions through an interaction mostly based on natural language utterances. However, although DAs are able to complete tasks such as sending texts, making phone calls, or playing songs, they still remain at an early stage in terms of their recommendation capabilities via a conversation. In addition, we have been witnessing the advent of increasingly precise and powerful recommendation algorithms and techniques able to effectively assess users’ tastes and predict information that may be of interest to them. Most of these approaches rely on the collaborative paradigm (often exploiting machine learning techniques) and neglect the huge amount of knowledge, both structured and unstructured, describing the domain of interest of a recommendation engine. Although very effective in predicting relevant items, collaborative approaches miss some very interesting features that go beyond the accuracy of results and move in the direction of providing novel and diverse results as well as generating explanations for recommended items. Knowledge-aware side information becomes crucial when a conversational interaction is implemented, in particular for preference elicitation, explanation, and critiquing steps.","PeriodicalId":443279,"journal":{"name":"Proceedings of the 16th ACM Conference on Recommender Systems","volume":"56 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115834856","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Siyong Ma, Puja Das, S. M. Nikolakaki, Qifeng Chen, Humeyra Topcu Altintas
Online commercial app marketplaces serve millions of apps to billions of users in an efficient manner. Bandit optimization algorithms are used to ensure that the recommendations are relevant, and converge to the best performing content over time. However, directly applying bandits to real-world systems, where the catalog of items is dynamic and continuously refreshed, is not straightforward. One of the challenges we face is the existence of several competing content surfacing components, a phenomenon not unusual in large-scale recommender systems. This often leads to challenging scenarios, where improving the recommendations in one component can lead to performance degradation of another, i.e., “cannibalization”. To address this problem we introduce an efficient two-layer bandit approach which is contextualized to user cohorts of similar taste. We mitigate cannibalization at runtime within a single multi-intent content surfacing platform by formalizing relevant offline evaluation metrics, and by involving the cross-component interactions in the bandit rewards. The user engagement in our proposed system has more than doubled as measured by online A/B testings.
{"title":"Two-Layer Bandit Optimization for Recommendations","authors":"Siyong Ma, Puja Das, S. M. Nikolakaki, Qifeng Chen, Humeyra Topcu Altintas","doi":"10.1145/3523227.3547396","DOIUrl":"https://doi.org/10.1145/3523227.3547396","url":null,"abstract":"Online commercial app marketplaces serve millions of apps to billions of users in an efficient manner. Bandit optimization algorithms are used to ensure that the recommendations are relevant, and converge to the best performing content over time. However, directly applying bandits to real-world systems, where the catalog of items is dynamic and continuously refreshed, is not straightforward. One of the challenges we face is the existence of several competing content surfacing components, a phenomenon not unusual in large-scale recommender systems. This often leads to challenging scenarios, where improving the recommendations in one component can lead to performance degradation of another, i.e., “cannibalization”. To address this problem we introduce an efficient two-layer bandit approach which is contextualized to user cohorts of similar taste. We mitigate cannibalization at runtime within a single multi-intent content surfacing platform by formalizing relevant offline evaluation metrics, and by involving the cross-component interactions in the bandit rewards. The user engagement in our proposed system has more than doubled as measured by online A/B testings.","PeriodicalId":443279,"journal":{"name":"Proceedings of the 16th ACM Conference on Recommender Systems","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132742521","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Naime Ranjbar Kermany, L. Pizzato, Thireindar Min, Callum Scott, A. Leontjeva
Australia’s largest bank, Commonwealth Bank (CBA) has a large data and analytics function that focuses on building a brighter future for all using data and decision science. In this work, we focus on creating better services for CBA customers by developing a next generation recommender system that brings the most relevant merchant reward offers that can help customers save money. Our recommender provides CBA cardholders with cashback offers from merchants, who have different objectives when they create offers. This work describes a multi-stakeholder, multi-objective problem in the context of CommBank Rewards (CBR) and describes how we developed a system that balances the objectives of the bank, its customers, and the many objectives from merchants into a single recommender system.
{"title":"A Multi-Stakeholder Recommender System for Rewards Recommendations","authors":"Naime Ranjbar Kermany, L. Pizzato, Thireindar Min, Callum Scott, A. Leontjeva","doi":"10.1145/3523227.3547388","DOIUrl":"https://doi.org/10.1145/3523227.3547388","url":null,"abstract":"Australia’s largest bank, Commonwealth Bank (CBA) has a large data and analytics function that focuses on building a brighter future for all using data and decision science. In this work, we focus on creating better services for CBA customers by developing a next generation recommender system that brings the most relevant merchant reward offers that can help customers save money. Our recommender provides CBA cardholders with cashback offers from merchants, who have different objectives when they create offers. This work describes a multi-stakeholder, multi-objective problem in the context of CommBank Rewards (CBR) and describes how we developed a system that balances the objectives of the bank, its customers, and the many objectives from merchants into a single recommender system.","PeriodicalId":443279,"journal":{"name":"Proceedings of the 16th ACM Conference on Recommender Systems","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114762248","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Marjan Celikik, Ana Peleteiro-Ramallo, Jacek Wasilewski
A large number of empirical studies on applying self-attention models in the domain of recommender systems are based on offline evaluation and metrics computed on standardized datasets. Moreover, many of them do not consider side information such as item and customer metadata although deep-learning recommenders live up to their full potential only when numerous features of heterogeneous type are included. Also, normally the model is used only for a single use case. Due to these shortcomings, even if relevant, previous works are not always representative of their actual effectiveness in real-world industry applications. In this talk, we contribute to bridging this gap by presenting live experimental results demonstrating improvements in user retention of up to 30%. Moreover, we share our learnings and challenges from building a re-usable and configurable recommender system for various applications from the fashion industry. In particular, we focus on fashion inspiration use-cases, such as outfit ranking, outfit recommendation and real-time personalized outfit generation.
{"title":"Reusable Self-Attention Recommender Systems in Fashion Industry Applications","authors":"Marjan Celikik, Ana Peleteiro-Ramallo, Jacek Wasilewski","doi":"10.1145/3523227.3547377","DOIUrl":"https://doi.org/10.1145/3523227.3547377","url":null,"abstract":"A large number of empirical studies on applying self-attention models in the domain of recommender systems are based on offline evaluation and metrics computed on standardized datasets. Moreover, many of them do not consider side information such as item and customer metadata although deep-learning recommenders live up to their full potential only when numerous features of heterogeneous type are included. Also, normally the model is used only for a single use case. Due to these shortcomings, even if relevant, previous works are not always representative of their actual effectiveness in real-world industry applications. In this talk, we contribute to bridging this gap by presenting live experimental results demonstrating improvements in user retention of up to 30%. Moreover, we share our learnings and challenges from building a re-usable and configurable recommender system for various applications from the fashion industry. In particular, we focus on fashion inspiration use-cases, such as outfit ranking, outfit recommendation and real-time personalized outfit generation.","PeriodicalId":443279,"journal":{"name":"Proceedings of the 16th ACM Conference on Recommender Systems","volume":"123 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116259294","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A users’ preference for a bundle – a set of items that can be purchased together – can be expressed by the utility of this bundle to the user. The multi-attribute utility theory motivate us to characterize the utility of a bundle using its attributes to improve the personalized bundle recommendation systems. This extended abstract for the Doctoral Symposium describes my PhD project for studying the utility of a bundle using its attributes. The steps taken and some preliminary results are presented, with an outline of the future plans.
{"title":"An Interpretable Neural Network Model for Bundle Recommendations: Doctoral Symposium, Extended Abstract","authors":"Xinyi Li, E. Malthouse","doi":"10.1145/3523227.3547423","DOIUrl":"https://doi.org/10.1145/3523227.3547423","url":null,"abstract":"A users’ preference for a bundle – a set of items that can be purchased together – can be expressed by the utility of this bundle to the user. The multi-attribute utility theory motivate us to characterize the utility of a bundle using its attributes to improve the personalized bundle recommendation systems. This extended abstract for the Doctoral Symposium describes my PhD project for studying the utility of a bundle using its attributes. The steps taken and some preliminary results are presented, with an outline of the future plans.","PeriodicalId":443279,"journal":{"name":"Proceedings of the 16th ACM Conference on Recommender Systems","volume":"168 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115079313","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：