{"corpus_id":221200085,"paper_sha":"af270108729de5d16d85a2d5ab77a0a8ac86910c","doi":"10.1109/tnnls.2020.3011559","arxiv_id":null,"pmid":32813663,"pmcid":null,"mag_id":3052825600,"dblp_id":"journals/tnn/WanHBRDDD21","acl_id":null,"title":"Human-in-the-Loop Low-Shot Learning","year":2020,"publication_date":"2020-08-19","venue":"IEEE Transactions on Neural Networks and Learning Systems","journal":{"name":"IEEE Transactions on Neural Networks and Learning Systems","pages":"3287-3292","volume":"32"},"journal_issn":null,"journal_title":null,"publication_types":["JournalArticle"],"pubmed_pub_types":["Journal Article","Research Support, Non-U.S. Gov't"],"s2_fields_of_study":["Medicine","Computer Science"],"reference_count":0,"citation_count":16,"influential_citation_count":0,"is_open_access":false,"arxiv_categories":null,"arxiv_license":null,"arxiv_journal_ref":null,"mesh_headings":[{"d":"Algorithms","mj":false,"ui":"D000465"},{"d":"Feedback","mj":false,"ui":"D005246"},{"d":"Humans","mj":false,"ui":"D006801"},{"d":"Learning","mj":false,"qs":[{"q":"physiology","mj":true,"ui":"Q000502"}],"ui":"D007858"},{"d":"Machine Learning","mj":true,"ui":"D000069550"},{"d":"Neural Networks, Computer","mj":false,"ui":"D016571"},{"d":"Problem Solving","mj":false,"ui":"D011340"},{"d":"Reinforcement, Psychology","mj":false,"ui":"D012054"},{"d":"Uncertainty","mj":false,"ui":"D035501"}],"chemicals":null,"comments_corrections":null,"source_flags":5,"s2_open_access_pdf_url":null,"s2_open_access_landing_url":null,"s2_open_access_license":null,"s2_open_access_status":null,"pmc_open_access_pdf_url":null,"pmc_open_access_landing_url":null,"pmc_open_access_license":null,"pmc_open_access_status":null,"unpaywall_open_access_pdf_url":null,"unpaywall_open_access_landing_url":null,"unpaywall_open_access_license":null,"unpaywall_open_access_status":null,"abstract":"We consider a human-in-the-loop scenario in the context of low-shot learning. Our approach was inspired by the fact that the viability of samples in novel categories cannot be sufficiently reflected by those limited observations. Some heterogeneous samples that are quite different from existing labeled novel data can inevitably emerge in the testing phase. To this end, we consider augmenting an uncertainty assessment module into low-shot learning system to account into the disturbance of those out-of-distribution (OOD) samples. Once detected, these OOD samples are passed to human beings for active labeling. Due to the discrete nature of this uncertainty assessment process, the whole Human-In-the-Loop Low-shot (HILL) learning framework is not end-to-end trainable. We hence revisited the learning system from the aspect of reinforcement learning and introduced the REINFORCE algorithm to optimize model parameters via policy gradient. The whole system gains noticeable improvements over existing low-shot learning approaches.","claims":[{"public_id":"cl_fe68203b148b458af8bc674258da7b6f","status":"active","text":"Detected out-of-distribution samples are passed to human beings for active labeling.","confidence":0.95,"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/claims/cl_fe68203b148b458af8bc674258da7b6f"},{"public_id":"cl_5d5e172fa9ed3230eebbd670b89d81f6","status":"active","text":"Human-In-the-Loop Low-shot learning augments a low-shot learning system with an uncertainty assessment module to account for out-of-distribution samples.","confidence":0.94,"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/claims/cl_5d5e172fa9ed3230eebbd670b89d81f6"},{"public_id":"cl_4908902cf270dd1e3932c3414379fd2d","status":"active","text":"REINFORCE optimizes model parameters in the framework via policy gradient after the learning system is revisited from a reinforcement learning perspective.","confidence":0.92,"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/claims/cl_4908902cf270dd1e3932c3414379fd2d"},{"public_id":"cl_2ff85ba27861a37f8fc1317116ab1ee3","status":"active","text":"The discrete uncertainty assessment process makes the Human-In-the-Loop Low-shot learning framework not end-to-end trainable.","confidence":0.9,"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/claims/cl_2ff85ba27861a37f8fc1317116ab1ee3"},{"public_id":"cl_5b383a0933a3ab50413afc8db9ae60ca","status":"active","text":"The whole system gains noticeable improvements over existing low-shot learning approaches.","confidence":0.86,"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/claims/cl_5b383a0933a3ab50413afc8db9ae60ca"}],"concepts":[{"public_id":"co_08bb051d0f4028bf56d95eb5eb3732d5","status":"active","name":"low-shot learning","description":"A learning setting where novel categories have only limited labeled observations.","types":["learning setting"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_08bb051d0f4028bf56d95eb5eb3732d5"},{"public_id":"co_19f750eed7320c014dea10689ae93e5b","status":"active","name":"human beings","description":"Human labelers who receive detected out-of-distribution samples for labeling.","types":["annotator group"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_19f750eed7320c014dea10689ae93e5b"},{"public_id":"co_2c9ce611fc3b947e5c2ec1969a84a77f","status":"active","name":"existing low-shot learning approaches","description":"Prior low-shot learning methods used as comparison targets for the reported improvements.","types":["baseline group"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_2c9ce611fc3b947e5c2ec1969a84a77f"},{"public_id":"co_320c760d71f21a745b69e2b664e85c9e","status":"active","name":"Human-In-the-Loop Low-shot learning framework","description":"A low-shot learning framework that incorporates human involvement for labeling uncertain samples.","types":["framework"],"aliases":["HILL"],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_320c760d71f21a745b69e2b664e85c9e"},{"public_id":"co_4207bf7c6085fdeb02b60564ed6497f1","status":"active","name":"heterogeneous samples","description":"Testing-phase samples that differ substantially from the existing labeled novel data.","types":["sample category"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_4207bf7c6085fdeb02b60564ed6497f1"},{"public_id":"co_425af14f97ce8ac70507f4c7bab98e10","status":"active","name":"reinforcement learning","description":"An optimization perspective used to revisit the low-shot learning system.","types":["learning paradigm"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_425af14f97ce8ac70507f4c7bab98e10"},{"public_id":"co_4373c1dfe90670908451edfb03e60ccc","status":"active","name":"REINFORCE algorithm","description":"A reinforcement learning algorithm introduced to optimize the framework through policy gradients.","types":["algorithm"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_4373c1dfe90670908451edfb03e60ccc"},{"public_id":"co_488e6b7cec3506ea947a3f1f18c35250","status":"active","name":"uncertainty assessment module","description":"A component added to the low-shot system to assess uncertainty from out-of-distribution disturbances.","types":["model component"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_488e6b7cec3506ea947a3f1f18c35250"},{"public_id":"co_6dde270e08b85e4fa232084684e180f6","status":"active","name":"out-of-distribution samples","description":"Samples treated as outside the expected distribution of the available labeled novel data.","types":["sample category"],"aliases":["OOD samples"],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_6dde270e08b85e4fa232084684e180f6"},{"public_id":"co_70003df0a9cf5e2987c01df5d15265c1","status":"active","name":"existing labeled novel data","description":"The available labeled examples for novel categories used by the low-shot system.","types":["training data"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_70003df0a9cf5e2987c01df5d15265c1"},{"public_id":"co_9d9b9a8c8511d2d22538a0cea71426d7","status":"active","name":"uncertainty assessment process","description":"The discrete procedure used to assess uncertainty in the human-in-the-loop framework.","types":["process"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_9d9b9a8c8511d2d22538a0cea71426d7"},{"public_id":"co_a289ad1528936a602bce1044e654d51e","status":"active","name":"novel categories","description":"New categories represented by limited observations in the low-shot learning setting.","types":["category group"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_a289ad1528936a602bce1044e654d51e"},{"public_id":"co_b24bd8e92385db148706d00956c1ca84","status":"active","name":"policy gradient","description":"The gradient-based optimization mechanism used with REINFORCE for model parameters.","types":["optimization method"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_b24bd8e92385db148706d00956c1ca84"},{"public_id":"co_b5f9a4e4e49453440d78e540fb18e10b","status":"active","name":"active labeling","description":"A labeling process in which selected uncertain samples are sent to humans for annotation.","types":["labeling process"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_b5f9a4e4e49453440d78e540fb18e10b"},{"public_id":"co_ebd82d3236836e48bc97fc940e17784c","status":"active","name":"model parameters","description":"The learnable quantities optimized in the human-in-the-loop low-shot learning framework.","types":["model component"],"aliases":[],"contributors":[{"id":35,"public_id":"b2adb6bfad","public_label":"Anonymous (b2adb6bfad)","roles":["extraction"],"url":"https://sah.borca.ai/u/b2adb6bfad"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"}],"url":"https://sah.borca.ai/concepts/co_ebd82d3236836e48bc97fc940e17784c"}],"external_ids":{"DOI":"10.1109/tnnls.2020.3011559","ArXiv":null,"PubMed":32813663,"PubMedCentral":null,"MAG":3052825600,"DBLP":"journals/tnn/WanHBRDDD21","ACL":null},"open_access":{"is_open_access":false,"pdf_url":null,"landing_url":"https://sah.borca.ai/papers/221200085","source":null,"pdf_url_source":null,"license":null,"reason":"pdf_url_not_indexed"},"reference_availability":{"status":"unknown","references_indexed":false,"full_text_available":false,"full_text_source":null,"count_basis":"semantic_scholar_metadata","extraction_status":"not_applicable","reason":null},"source":{"provider":"episteme2","base_corpus":"semantic_scholar_dump","freshness_mode":"unknown","basis":["semantic_scholar_metadata","postgres_metadata"],"limits":["paper metadata is based on indexed upstream scholarly datasets","claims and concepts are available only for extracted papers","absence of claims or concepts means no extracted graph data is available in this response"],"status":"available","degraded":false,"degraded_reasons":[],"diagnostics":{"status":"available","degraded":false,"degraded_reasons":[],"metadata_status":"available","graph_status":"available","abstract_status":"available"},"source_flags":5},"paper_id":631924,"paper_uid":"9c4708ba-11de-4d85-b3b5-836acdde8090","canonical_identity":{"paper_id":631924,"paper_uid":"9c4708ba-11de-4d85-b3b5-836acdde8090","identity_status":"available","lookup_basis":"semantic_scholar_external_id","compatibility_path":"corpus_id"},"url":"https://sah.borca.ai/papers/221200085"}