{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/compatibility-family-learning-for-item","title":"Compatibility Family Learning for Item Recommendation and Generation","arxiv_id":"1712.01262","date":"2017-12-02","proceeding":null,"authors":["Yong-Siang Shih","Kai-Yueh Chang","Hsuan-Tien Lin","Min Sun"],"abstract":"Compatibility between items, such as clothes and shoes, is a major factor\namong customer's purchasing decisions. However, learning \"compatibility\" is\nchallenging due to (1) broader notions of compatibility than those of\nsimilarity, (2) the asymmetric nature of compatibility, and (3) only a small\nset of compatible and incompatible items are observed. We propose an end-to-end\ntrainable system to embed each item into a latent vector and project a query\nitem into K compatible prototypes in the same space. These prototypes reflect\nthe broad notions of compatibility. We refer to both the embedding and\nprototypes as \"Compatibility Family\". In our learned space, we introduce a\nnovel Projected Compatibility Distance (PCD) function which is differentiable\nand ensures diversity by aiming for at least one prototype to be close to a\ncompatible item, whereas none of the prototypes are close to an incompatible\nitem. We evaluate our system on a toy dataset, two Amazon product datasets, and\nPolyvore outfit dataset. Our method consistently achieves state-of-the-art\nperformance. Finally, we show that we can visualize the candidate compatible\nprototypes using a Metric-regularized Conditional Generative Adversarial\nNetwork (MrCGAN), where the input is a projected prototype and the output is a\ngenerated image of a compatible item. We ask human evaluators to judge the\nrelative compatibility between our generated images and images generated by\nCGANs conditioned directly on query items. Our generated images are\nsignificantly preferred, with roughly twice the number of votes as others.","url_abs":"http://arxiv.org/abs/1712.01262v1","url_pdf":"http://arxiv.org/pdf/1712.01262v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"compatibility-family-learning-for-item","repo_url":"https://github.com/appier/compatibility-family-learning","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":null,"task_name":"Generative Adversarial Network"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1712.01262","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}