?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=Learning+Stochastic+Majority+Votes+by+Minimizing+a+PAC-Bayes+Generalization+Bound&rft.creator=Zantedeschi%2C+V&rft.creator=Viallard%2C+P&rft.creator=Morvant%2C+E&rft.creator=Emonet%2C+R&rft.creator=Habrard%2C+A&rft.creator=Germain%2C+P&rft.creator=Guedj%2C+B&rft.description=We+investigate+a+stochastic+counterpart+of+majority+votes+over+finite+ensembles+of+classifiers%2C+and+study+its+generalization+properties.+While+our+approach+holds+for+arbitrary+distributions%2C+we+instantiate+it+with+Dirichlet+distributions%3A+this+allows+for+a+closed-form+and+differentiable+expression+for+the+expected+risk%2C+which+then+turns+the+generalization+bound+into+a+tractable+training+objective.The+resulting+stochastic+majority+vote+learning+algorithm+achieves+state-of-the-art+accuracy+and+benefits+from+(non-vacuous)+tight+generalization+bounds%2C+in+a+series+of+numerical+experiments+when+compared+to+competing+algorithms+which+also+minimize+PAC-Bayes+objectives+--+both+with+uninformed+(data-independent)+and+informed+(data-dependent)+priors.&rft.publisher=Advances+in+Neural+Information+Processing+Systems+(NeurIPS+2021)&rft.date=2021&rft.type=Proceedings+paper&rft.language=eng&rft.source=+++++In%3A++Proceedings+of+the+35th+Conference+on+Neural+Information+Processing+Systems+(NeurIPS+2021).++++Advances+in+Neural+Information+Processing+Systems+(NeurIPS+2021)+(2021)+++++&rft.format=text&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10138174%2F7%2FGuedj_Learning%2520Stochastic%2520Majority%2520Votes%2520by%2520Minimizing%2520a%2520PAC-Bayes%2520Generalization%2520Bound_VoR.pdf&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10138174%2F&rft.rights=open