Choosing the Sample with Lowest Loss makes SGD Robust

Shah Vatsal
Shah Vatsal
Wu Xiaoxia
Wu Xiaoxia

AISTATS, pp. 2120-2130, 2020.

Cited by: 0|Bibtex|Views7|Links
EI

Abstract:

The presence of outliers can potentially significantly skew the parameters of machine learning models trained via stochastic gradient descent (SGD). In this paper we propose a simple variant of the simple SGD method: in each step, first choose a set of k samples, then from these choose the one with the smallest current loss, and do an S...More

Code:

Data:

Your rating :
0

 

Tags
Comments