?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=Probabilistic+Active+Meta-Learning&rft.creator=Kaddour%2C+J&rft.creator=S%C3%A6mundsson%2C+S&rft.creator=Deisenroth%2C+MP&rft.description=Data-efficient+learning+algorithms+are+essential+in+many+practical+applications+where+data+collection+is+expensive%2C+e.g.%2C+in+robotics+due+to+the+wear+and+tear.+To+address+this+problem%2C+meta-learning+algorithms+use+prior+experience+about+tasks+to+learn+new%2C+related+tasks+efficiently.+Typically%2C+a+set+of+training+tasks+is+assumed+given+or+randomly+chosen.+However%2C+this+setting+does+not+take+into+account+the+sequential+nature+that+naturally+arises+when+training+a+model+from+scratch+in+real-life%3A+how+do+we+collect+a+set+of+training+tasks+in+a+data-efficient+manner%3F+In+this+work%2C+we+introduce+task+selection+based+on+prior+experience+into+a+meta-learning+algorithm+by+conceptualizing+the+learner+and+the+active+meta-learning+setting+using+a+probabilistic+latent+variable+model.+We+provide+empirical+evidence+that+our+approach+improves+data-efficiency+when+compared+to+strong+baselines+on+simulated+robotic+experiments.&rft.publisher=NeurIPS+Proceedings&rft.contributor=Larochelle%2C+H&rft.contributor=Ranzato%2C+M&rft.contributor=Hadsell%2C+R&rft.contributor=Balcan%2C+M-F&rft.contributor=Lin%2C+H-T&rft.date=2020-12-12&rft.type=Proceedings+paper&rft.language=eng&rft.source=+++++In%3A+Larochelle%2C+H+and+Ranzato%2C+M+and+Hadsell%2C+R+and+Balcan%2C+M-F+and+Lin%2C+H-T%2C+(eds.)+Advances+in+Neural+Information+Processing+Systems+33+pre-proceedings+(NeurIPS+2020).++++NeurIPS+Proceedings%3A+Virtual+conference.+(2020)++++(In+press).++&rft.format=text&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10120832%2F1%2FNeurIPS-2020-probabilistic-active-meta-learning-Paper.pdf&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10120832%2F&rft.rights=open