Abstract
Autoencoder node saliency (ANS) is a novel method intended to explain the lenrning tasks perfonned in the hidden layer of an autoencoder. ANS ranks the hidden nodes based on their ability to perfonn a given learning task. It identifies the hidden node that best distinguishes between two classes of data points. The weights from the identified hidden nodes correspond to original features that contribute to the data differentiation.
- Developers:
-
Fan, Ya [1]
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Release Date:
- 2017-11-21
- Project Type:
- Open Source, Publicly Available Repository
- Software Type:
- Scientific
- Licenses:
-
MIT License
- Sponsoring Org.:
-
USDOE National Nuclear Security Administration (NNSA)Primary Award/Contract Number:AC52-07NA27344
- Code ID:
- 16356
- Site Accession Number:
- LLNL-CODE-753346
- Research Org.:
- Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)
- Country of Origin:
- United States
Citation Formats
Fan, Ya J.
Autoencoder Node Saliency.
Computer Software.
https://github.com/LLNL/ANS.
USDOE National Nuclear Security Administration (NNSA).
21 Nov. 2017.
Web.
doi:10.11578/dc.20180731.9.
Fan, Ya J.
(2017, November 21).
Autoencoder Node Saliency.
[Computer software].
https://github.com/LLNL/ANS.
https://doi.org/10.11578/dc.20180731.9.
Fan, Ya J.
"Autoencoder Node Saliency." Computer software.
November 21, 2017.
https://github.com/LLNL/ANS.
https://doi.org/10.11578/dc.20180731.9.
@misc{
doecode_16356,
title = {Autoencoder Node Saliency},
author = {Fan, Ya J.},
abstractNote = {Autoencoder node saliency (ANS) is a novel method intended to explain the lenrning tasks perfonned in the hidden layer of an autoencoder. ANS ranks the hidden nodes based on their ability to perfonn a given learning task. It identifies the hidden node that best distinguishes between two classes of data points. The weights from the identified hidden nodes correspond to original features that contribute to the data differentiation.},
doi = {10.11578/dc.20180731.9},
url = {https://doi.org/10.11578/dc.20180731.9},
howpublished = {[Computer Software] \url{https://doi.org/10.11578/dc.20180731.9}},
year = {2017},
month = {nov}
}