Stable Extraction of Frequent Sub-sequences from Sequential Symbol Input
Kenta Morita, Haruhiko Takase, Naoki Morita, Hiroharu Kawanaka, Shinji Tsuruoka · Procedia Computer Science · 2017
In this paper, we discuss the method for stable extraction of frequent sub-sequences from symbols infused continuously. To make learning algorithm stably, we focus on two problems: the case where input data contain plural frequent sub-sequences and the case where appearance rate of frequent sub-sequences is low. To overcome these problems, we propose two ideas: new incremental structure and new learning rule. By simple experiments, we show the effectiveness of the proposed method.