{"id":"https://openalex.org/W2547875792","doi":"https://doi.org/10.48550/arxiv.1611.01144","title":"Categorical Reparameterization with Gumbel-Softmax","display_name":"Categorical Reparameterization with Gumbel-Softmax","publication_year":2016,"publication_date":"2016-11-03","ids":{"openalex":"https://openalex.org/W2547875792","doi":"https://doi.org/10.48550/arxiv.1611.01144","mag":"2547875792"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:1611.01144","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/1611.01144","pdf_url":"https://arxiv.org/pdf/1611.01144","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/1611.01144","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5083516691","display_name":"Eric Jang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jang, Eric","raw_affiliation_strings":["Google Brain"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Brain","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5061613634","display_name":"Shixiang Gu","orcid":"https://orcid.org/0000-0001-9246-0896","observed_orcids":["https://orcid.org/0000-0001-9246-0896"]},"institutions":[{"id":"https://openalex.org/I241749","display_name":"University of Cambridge","ror":"https://ror.org/013meh722","country_code":"GB","type":"education","lineage":["https://openalex.org/I241749"]},{"id":"https://openalex.org/I8087733","display_name":"University of T\u00fcbingen","ror":"https://ror.org/03a1kwz48","country_code":"DE","type":"education","lineage":["https://openalex.org/I8087733"]}],"countries":["DE","GB"],"is_corresponding":false,"raw_author_name":"Gu, Shixiang","raw_affiliation_strings":["University of Cambridge MPI T\u00fcbingen"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Cambridge MPI T\u00fcbingen","institution_ids":["https://openalex.org/I241749","https://openalex.org/I8087733"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5004918186","display_name":"Ben Poole","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I97018004","display_name":"Stanford University","ror":"https://ror.org/00f54p054","country_code":"US","type":"education","lineage":["https://openalex.org/I97018004"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Poole, Ben","raw_affiliation_strings":["Stanford University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Stanford University","institution_ids":["https://openalex.org/I97018004"]}]}],"institutions":[],"countries_distinct_count":3,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":5403,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.2996000051498413,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.2996000051498413,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12814","display_name":"Gaussian Processes and Bayesian Inference","score":0.16259999573230743,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10320","display_name":"Neural Networks and Applications","score":0.11020000278949738,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/categorical-latent-variables","display_name":"categorical latent variables","score":0.7390000224113464},{"id":"https://openalex.org/keywords/differentiable-sampling","display_name":"differentiable sampling","score":0.8309999704360962},{"id":"https://openalex.org/keywords/gradient-estimation","display_name":"gradient estimation","score":0.746999979019165},{"id":"https://openalex.org/keywords/annealing","display_name":"annealing","score":0.5090000033378601},{"id":"https://openalex.org/keywords/structured-prediction","display_name":"structured prediction","score":0.8920000195503235},{"id":"https://openalex.org/keywords/semi-supervised-classification","display_name":"semi-supervised classification","score":0.9229999780654907},{"id":"https://openalex.org/keywords/stochastic-neural-networks","display_name":"stochastic neural networks","score":0.7900000214576721}],"concepts":[{"id":"https://openalex.org/C5274069","wikidata":"https://www.wikidata.org/wiki/Q2285707","display_name":"Categorical variable","level":2,"score":0.8661178350448608},{"id":"https://openalex.org/C137610916","wikidata":"https://www.wikidata.org/wiki/Q1096862","display_name":"Gumbel distribution","level":3,"score":0.7950888872146606},{"id":"https://openalex.org/C188441871","wikidata":"https://www.wikidata.org/wiki/Q7554146","display_name":"Softmax function","level":3,"score":0.7324645519256592},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.6358586549758911},{"id":"https://openalex.org/C51167844","wikidata":"https://www.wikidata.org/wiki/Q4422623","display_name":"Latent variable","level":2,"score":0.5904943943023682},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5517497062683105},{"id":"https://openalex.org/C67926830","wikidata":"https://www.wikidata.org/wiki/Q5051810","display_name":"Categorical distribution","level":5,"score":0.5146666169166565},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.500476598739624},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.402866929769516},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.3723059594631195},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.37099844217300415},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.3354898691177368},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.27210256457328796},{"id":"https://openalex.org/C147581598","wikidata":"https://www.wikidata.org/wiki/Q729429","display_name":"Extreme value theory","level":2,"score":0.06533408164978027},{"id":"https://openalex.org/C160947583","wikidata":"https://www.wikidata.org/wiki/Q2083147","display_name":"Distribution fitting","level":3,"score":0.0},{"id":"https://openalex.org/C57205106","wikidata":"https://www.wikidata.org/wiki/Q3258519","display_name":"Inverse-chi-squared distribution","level":4,"score":0.0}],"mesh":[],"locations_count":3,"locations":[{"id":"pmh:oai:arXiv.org:1611.01144","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/1611.01144","pdf_url":"https://arxiv.org/pdf/1611.01144","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"mag:2950151997","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/1611.01144.pdf","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null},{"id":"doi:10.48550/arxiv.1611.01144","doi":"https://doi.org/10.48550/arxiv.1611.01144","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.1611.01144","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:1611.01144","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/1611.01144","pdf_url":"https://arxiv.org/pdf/1611.01144","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W2547875792.pdf","grobid_xml":"https://content.openalex.org/works/W2547875792.grobid-xml"},"referenced_works_count":22,"referenced_works":["https://openalex.org/W10548402","https://openalex.org/W1481935768","https://openalex.org/W1514535095","https://openalex.org/W1646152356","https://openalex.org/W1909320841","https://openalex.org/W1959608418","https://openalex.org/W2046765929","https://openalex.org/W2096192494","https://openalex.org/W2097268041","https://openalex.org/W2119717200","https://openalex.org/W2122262818","https://openalex.org/W2135181320","https://openalex.org/W2242818861","https://openalex.org/W2259167532","https://openalex.org/W2278112683","https://openalex.org/W2296701362","https://openalex.org/W2510842514","https://openalex.org/W2530887700","https://openalex.org/W2949416428","https://openalex.org/W2949595773","https://openalex.org/W2949605076","https://openalex.org/W2951004968"],"related_works":["https://openalex.org/W7119066800","https://openalex.org/W4388962827","https://openalex.org/W3010148210","https://openalex.org/W4280628756","https://openalex.org/W3208398914","https://openalex.org/W2943262294","https://openalex.org/W3128295440","https://openalex.org/W3168771896","https://openalex.org/W4415337197","https://openalex.org/W3106085430"],"abstract_inverted_index":{"Categorical":[0],"variables":[1,21],"are":[2],"a":[3,45,49,53,70],"natural":[4],"choice":[5],"for":[6],"representing":[7],"discrete":[8],"structure":[9],"in":[10],"the":[11,24,41,60],"world.":[12],"However,":[13],"stochastic":[14],"neural":[15],"networks":[16],"rarely":[17],"use":[18],"categorical":[19,46,71,93],"latent":[20,94],"due":[22],"to":[23,26],"inability":[25],"backpropagate":[27],"through":[28],"samples.":[29],"In":[30],"this":[31],"work,":[32],"we":[33],"present":[34],"an":[35],"efficient":[36],"gradient":[37,81],"estimator":[38,78],"that":[39,63,75],"replaces":[40],"non-differentiable":[42],"sample":[43,51],"from":[44,52],"distribution":[47,58],"with":[48,92],"differentiable":[50],"novel":[54],"Gumbel-Softmax":[55,77],"distribution.":[56,72],"This":[57],"has":[59],"essential":[61],"property":[62],"it":[64],"can":[65],"be":[66],"smoothly":[67],"annealed":[68],"into":[69],"We":[73],"show":[74],"our":[76],"outperforms":[79],"state-of-the-art":[80],"estimators":[82],"on":[83,100],"structured":[84],"output":[85],"prediction":[86],"and":[87,96],"unsupervised":[88],"generative":[89],"modeling":[90],"tasks":[91],"variables,":[95],"enables":[97],"large":[98],"speedups":[99],"semi-supervised":[101],"classification.":[102]},"counts_by_year":[{"year":2026,"cited_by_count":450},{"year":2025,"cited_by_count":582},{"year":2024,"cited_by_count":792},{"year":2023,"cited_by_count":753},{"year":2022,"cited_by_count":662},{"year":2021,"cited_by_count":814},{"year":2020,"cited_by_count":612},{"year":2019,"cited_by_count":430},{"year":2018,"cited_by_count":227},{"year":2017,"cited_by_count":77},{"year":2016,"cited_by_count":4}],"updated_date":"2026-11-12T12:28:57.059296","created_date":"2016-11-11T00:00:00"}