Summary
2011 International Symposium on Nonlinear Theory and Its Applications
2011
Session Number:A2L-B
Session:
Number:A2L-B3
Performance Enhancement of Amoeba-based Neurocomputer for 8-City Traveling Salesman Problem
Masashi Aono, Liping Zhu, Song-Ju Kim, Masahiko Hara,
pp.104-107
Publication Date:2011/9/4
Online ISSN:2188-5079
Summary:
Many biological organisms manage a tradeoff between the accuracy and speed of their decision making that are conflicted objectives for surviving in uncertain environments. A single-celled amoeboid organism, the true slime mold Physarum polycephalum, exhibits rich spatiotemporal oscillatory dynamics and sophisticated resource allocation capabilities. To evaluate the accuracy and speed of the resource allocation, previously the authors constructed an experimental computing system that leads the organism to search for a solution to the 8-city Traveling Salesman Problem (TSP).With the assistance of optical feedback to implement a recurrent neural network model, the organism changes its shape by alternately expanding and shrinking its photosensitive branches so that its body area can be maximized and the risk of being illuminated can be minimized. The system found a high quality TSP solution (i.e., a relatively shorter route) with a high accuracy. In this study, we show that, with changes in the experimental condition of the optical feedback system, the speed at which the system searches for the solution was enhanced significantly without degrading the accuracy. Counterintuitively, to speed up the search process, the organism acquired less information from the optical feedback compared with our previously performed experiments. These findings suggest that the search dynamics of the organism may be tuned to become "economical" to get a high quality solution at lower exploration cost.