Response Characterization for Auditing Cell Dynamics in Long Short-term Memory Networks
Name
1809.03864.pdf
Description
Submitted version
Size
1.97 MB
Format
Adobe PDF
Checksum (MD5)
f65f159923c7b9a0c411fd7a3d84badd
Author(s)
Hasani, Ramin
Amini, Alexander A
Lechner, Mathias
Naser, Felix M
Grosu, Radu
Rus, Daniela L
Date Issued
September 2019
Journal
2019 International Joint Conference on Neural Networks
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Citation
Hasani, Ramin et al. "Response Characterization for Auditing Cell Dynamics in Long Short-term Memory Networks." 2019 International Joint Conference on Neural Networks, July 2019, Budapest, Hungary, Institute of Electrical and Electronics Engineers, September 2019. © 2019 IEEE
Version
Original manuscript
Abstract
In this paper, we introduce a novel method to interpret recurrent neural networks (RNNs), particularly long short-term memory networks (LSTMs) at the cellular level. We propose a systematic pipeline for interpreting individual hidden state dynamics within the network using response characterization methods. The ranked contribution of individual cells to the network's output is computed by analyzing a set of interpretable metrics of their decoupled step and sinusoidal responses. As a result, our method is able to uniquely identify neurons with insightful dynamics, quantify relationships between dynamical properties and test accuracy through ablation analysis, and interpret the impact of network capacity on a network's dynamical distribution. Finally, we demonstrate the generalizability and scalability of our method by evaluating a series of different benchmark sequential datasets.
MIT Department
Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
Creative Commons Attribution-Noncommercial-Share Alike
Persistent DSpace Link
DOI of Published Version
https://doi.org/10.1109/IJCNN.2019.8851954
