Goto

Collaborating Authors

 Country






f4e3ce3e7b581ff32e40968298ba013d-Supplemental.pdf

Neural Information Processing Systems

We also discuss other relevant hyper-parameters. Finally, we provide additional results from the data domaindifferentthanimages. BackgroundandSetting For the convenience of the reader, we restate our notation here. We define two kinds of purity ofA on Sn. We first develop some notation. LetVd be the volume of the unit d-dimensional ball.


f4e3ce3e7b581ff32e40968298ba013d-Paper.pdf

Neural Information Processing Systems

Byleveraging thehigh-order topological information ofdata,weareable to collect most of the clean data and train a high-quality model. Theoretically we prove that this topological approach is guaranteed to collect the clean data with high probability.



c74214a3877c4d8297ac96217d5189b7-Paper.pdf

Neural Information Processing Systems

However, the resulting methods often suffer from high computational complexity which has reduced their practical applicability. For example, in the case of multiclass logistic regression, the aggregating forecaster (Foster et al. (2018)) achievesaregret ofO(log(Bn))whereas Online Newton Step achieves O(eBlog(n))obtaining adouble exponential gaininB (aboundonthenormof comparativefunctions).