One-sidedFrank-WolfealgorithmsforSaddleproblemsVladimirKolmogorov1ThomasPock2AbstractwhereX,Yarefinitedimensionalspaces,equippedwithaninnerproduct·,·andK:X→YisaboundedWestudyaclassofconvex-conca...
EscapingSaddlePointswithAdaptiveGradientMethodsMatthewStaib12SashankReddi3SatyenKale3SanjivKumar3SuvritSra1AbstractAdagradupdatestheparametersinthefollowingmanner:AdaptivemethodssuchasAdamandRMSPro...
DefendingAgainstSaddlePointAttackinByzantine-RobustDistributedLearningDongYin1YudongChen2KannanRamchandran1PeterBartlett13Abstractarestoredandprocessedlocallyinendusers’cellphonesandpersonalcomput...