Response Characterization for Auditing Cell Dynamics in Long Short-term Memory Networks

• Main Authors: Hasani, Ramin (Author), Amini, Alexander A (Author), Lechner, Mathias (Author), Naser, Felix M (Author), Grosu, Radu (Author), Rus, Daniela L (Author)
• Other Authors: Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory (Contributor), Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science (Contributor)
• Format: Article
• Language: English
• Published: Institute of Electrical and Electronics Engineers (IEEE), 2021-05-04T14:56:09Z.
• Subjects: Article
• Online Access: Get fulltext

 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.