[3K1-OS-5a] OS-5 (1)

[3K1-OS-5a-03] Extraction and Interpretation of Classification Patterns from Deep Learning Networks using HMM

〇Masayuki Ando¹, Yoshinobu Kawahara^2,3, Wataru Sunayama⁴, Yuji Hatanaka⁴ (1. Graduate School of Engineering, The University of Shiga Prefecture, 2. RIKEN Center for Advanced Intelligence Project, 3. Institute of Mathematics for Industry, Kyushu University, 4. School of Engineering, The University of Shiga Prefecture)

Keywords:Deep Learning, Text Mining, Interpretation Support

In deep learning, there is a problem that concrete classification patterns for deriving reasons for classification are often incomprehensible. In this paper, we propose a classification patterns extraction system from deep learning networks and verified the effectiveness of the system. The proposed system extracts classification patterns from the trained learning networks of LSTM using HMM. Then the system displays the extracted classification patterns so that users of the system can interpret the learning networks. In verification experiments, the frequency of the extracted classification patterns and the frequency of the classification patterns based on the frequency were compared. Then, We confirmed effectiveness of the interpretations of the extracted patterns by the subjects. The results showed that the proposed system can extract classification patterns effective for interpretations of the learning networks.