Conference Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 15 - Subsampling: Basic Tool That Facilitates the Identification of Statistical Relationships in Big Data
Type: Topic Contributed
Date/Time: Sunday, August 7, 2022 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #323331
Title: Using Subsampling to Speed up Training in Attention-Based NNs: A Case Study at Doing Statistics at Scale in the Amazon Supply Chain
Author(s): Dean p foster* and Kenny Shirley
Companies: Amazon and Amazon
Keywords:
Abstract:

Merely doing the simple task of evaluating a neural network on 20 million products can take several days on a fairly larger cluster. So when it comes to training and fitting, using subsampling techniques are important. We will discuss several ways that we have found subsampling useful in doing statistics on large data sets.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2022 program