Interpretable Recurrent Neural Networks Using Sequential Sparse Recovery

By Scott Wisdom, Thomas Powers, James W. Pitton, Les E. Atlas. 2016

37 citations recorded by the source corpus.

Recurrent neural networks (RNNs) are powerful and effective for processing sequential data. However, RNNs are usually considered "black box" models whose internal structure and learned parameters are not interpretable. In this paper, we propose an interpretable RNN based on the sequential iterative soft-thresholding algorithm (SISTA) for solving the sequential sparse recovery problem, which models a sequence of correlated observations with a sequence of sparse latent vectors. The architecture of the resulting SISTA-RNN is implicitly defined by the computational structure of SISTA, which results in a novel stacked RNN architecture. Furthermore, the weights of the SISTA-RNN are perfectly interpretable as the parameters of a principled statistical model, which in this case include a sparsifying dictionary, iterative step size, and regularization parameters. In addition, on a particular sequential compressive sensing task, the SISTA-RNN trains faster and achieves better performance than conventional state-of-the-art black box RNNs, including long-short term memory (LSTM) RNNs.

Subfields: Computational Mechanics, Computer Vision and Pattern Recognition, Model Architectures, Safety, Robustness & Interpretability, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mechanics

1. Yi Ma
2. John Wright
3. Shenghua Gao
4. Chong You
5. Dino Pedreschi
6. Weixin Luo
7. Wen Liu
8. Yaodong Yu

In Computer Vision and Pattern Recognition

1. Shenghua Gao
2. Lei Zhang
3. Ming-Hsuan Yang
4. Luc Van Gool
5. Dacheng Tao
6. Yi Ma
7. John Wright
8. Shuicheng Yan

In Model Architectures

1. Kwan Ho Ryan Chan
2. Shenghua Gao
3. Chong You
4. Yaodong Yu
5. Yi Ma
6. Haozhi Qi

In Safety, Robustness & Interpretability

1. Cecilia Panigutti
2. Dino Pedreschi
3. Alan Perotti
4. Yaodong Yu
5. Salvatore Ruggieri
6. Christian Walter Peter Omlin
7. Waddah Saeed
8. Surabhi Verma

In Signal Processing

1. Yi Ma
2. John Wright
3. Dino Pedreschi
4. Shenghua Gao
5. Chong You
6. Weixin Luo
7. Wen Liu
8. René Vidal

Citations in the corpus

• Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AI (2019, neutral): “To create an interpretable RNN, the authors in [283] pro-pose an RNN based on SISTA (Sequential Iterative Soft-Thresholding Algorithm) that models a sequence of correlated observations with a sequence of sparse latent vectors, making its weights interpretable as the parameters of a principled…”
• Deep Networks from the Principle of Rate Reduction (2020, methodology): “As suggested in Yu et al. (2020), the so learned features can be directly used for classification via a nearest subspace classifier.”
• A Revisit of Sparse Coding Based Anomaly Detection in Stacked RNN Framework (2017, methodology): “Mathematically, the stacked RNN (sRNN) can be written as follows [26]:”
• Doctor XAI: an ontology-based approach to black-box sequential data classification explanations (2020, background): “Other interpretability approaches related to sequential data modeling focus on understanding the internal behavior of the black-box under study [45]; conversely, our approach is agnostic to the black-box whose outcome we want to explain.”