EXPLORING THE USE OF LONG SHORT-TERM MEMORY (LSTM) IN FUNCTIONAL BASED BIOINSPIRED DESIGN
Year: 2018
Editor: Elies Dekoninck, Andrew Wodehouse, Chris Snider, Georgi Georgiev, Gaetano Cascini
Author: Jin Ma, Shuo Jiang, Jie Hu, Jian Shen, Jin Qi, Pan Tian
Series: ICDC
Section: BIO-INSPIRED CREATIVITY
Page(s): 338-345
ISBN: 9781912254071
Abstract
Bioinspired design is the rising discipline where biological cases are taken to inspire the designers to solve the engineering problems or challenges. The search of functional analogies using biological repository for analogies and inspirations acquisition based on engineering-to-biology thesaurus has proven to be an effective way. Whereas bias and inconsistency exist in the manually made thesaurus seriously hinders the exploration of biological cases. This research presents an Long Short-Term Memory (LSTM) based analogical reasoning methodology for bioinspired design, which focus on the realization of inter-disciplinary knowledge reasoning through extracting thesaurus from functional models automatically. Its effectiveness has been analyzed on an engineering-to-biology depository in terms of perplexity and functional models error rate. Furthermore, comparisons between the proposed method and other methods are given, and results show that considerable improvements are gained for analogical reasoning in functional based bioinspired design.
Keywords: Bioinspired Design, Functional Model, LSTM, Concept generation