Recurrent Neural Networks (RNNs) have demonstrated their effectiveness in learning and processing sequential data (e.g., speech and natural language). However, due to the black-box nature of neural networks, understanding the decision logic of RNNs is quite challenging. Some recent progress has been made to approximate the behavior of an RNN by weighted automata. They provide a better interpretability, but still suffer from poor scalability. In this paper, we propose a novel approach to extracting weighted automata with the guidance of a target RNN's decision and context information. In particular, we identify the patterns of its step-wise predictive decisions to instruct the formation of automata states. Further, we propose a state composition method to enhance the context-awareness of the extracted model. Our in-depth evaluations on typical RNN tasks, including language model and classification, demonstrate the effectiveness and advantage of our method over the state-of-the-arts. It achieves an accurate approximation of an RNN even on large-scale tasks.
Jul 13th 2021
Mar 4th 2021
Jan 22nd 2021
Looking to build AI capacity? Need a speaker at your event?
Get involved in Alberta's growing AI ecosystem! Speaker, sponsorship, and letter of support requests welcome.
Curious about study options under one of our researchers? Want more information on training opportunities?