?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=Online+Parameter-Free+Learning+of+Multiple+Low+Variance+Tasks&rft.creator=Denevi%2C+G&rft.creator=Pontil%2C+M&rft.creator=Stamos%2C+D&rft.description=We+propose+a+method+to+learn+a+common+bias+vector+for+a+growing+sequence+of+low-variance+tasks.+Unlike+state-of-the-art+approaches%2C+our+method+does+not+require+tuning+any+hyper-parameter.+Our+approach+is+presented+in+the+non-statistical+setting+and+can+be+of+two+variants.+The+%E2%80%9Caggressive%E2%80%9D+one+updates+the+bias+after+each+datapoint%2C+the+%E2%80%9Clazy%E2%80%9D+one+updates+the+bias+only+at+the+end+of+each+task.+We+derive+an+across-tasks+regret+bound+for+the+method.+When+compared+to+state-of-the-art+approaches%2C+the+aggressive+variant+returns+faster+rates%2C+the+lazy+one+recovers+standard+rates%2C+but+with+no+need+of+tuning+hyper-parameters.+We+then+adapt+the+methods+to+the+statistical+setting%3A+the+aggressive+variant+becomes+a+multi-task+learning+method%2C+the+lazy+one+a+meta-learning+method.+Experiments+confirm+the+effectiveness+of+our+methods+in+practice.&rft.publisher=AUAI+Press&rft.contributor=Adams%2C+RP&rft.contributor=Gogate%2C+V&rft.date=2020-08-03&rft.type=Proceedings+paper&rft.language=eng&rft.source=+++++In%3A+Adams%2C+RP+and+Gogate%2C+V%2C+(eds.)+Proceedings+of+the+36th+Conference+on+Uncertainty+in+Artificial+Intelligence+(UAI).++++AUAI+Press%3A+Virtual+conference.+(2020)+++++&rft.format=text&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10112125%2F1%2F368_main_paper.pdf&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F10112125%2F&rft.rights=open