|
Activity Number:
|
524
- Emerging Statistical Learning Methods in Modern Data Science
|
|
Type:
|
Invited
|
|
Date/Time:
|
Thursday, August 6, 2020 : 1:00 PM to 2:50 PM
|
|
Sponsor:
|
Section on Statistical Learning and Data Science
|
|
Abstract #309488
|
|
|
Title:
|
Reverse Engineering a Deep Network
|
|
Author(s):
|
Douglas Nychka*
|
|
Companies:
|
Colorado School of Mines
|
|
Keywords:
|
deep learning;
spatial processes
|
|
Abstract:
|
Deep learning networks have proved to be feasible to estimate based on training data sets and have had success in feature detection. It is also widely recognized that these networks are difficult to interpret. In this work we attempt to find a spatial statistical model that provides the same kind of performance as the network on training images. The idea is that this spatial model then becomes a stand-in for understanding what the network is sensitive to in detecting features. Developing a statistical surrogate also opens up the possibility of characterizing uncertainty from the network output using a more statistical framework.
|
Authors who are presenting talks have a * after their name.
