Euglena-based neurocomputing with two-dimensional optical feedback on swimming cells in micro-aquariums

Authors:
Kazunari Ozasa;Jeesoo Lee;Simon Song;Masahiko Hara;Mizuo Maeda
Affiliations:
RIKEN, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan;Department of Mechanical Engineering, Hanyang University, 17 Haendang-dong, Seongdong-gu, Seoul 133-791, Republic of Korea;Department of Mechanical Engineering, Hanyang University, 17 Haendang-dong, Seongdong-gu, Seoul 133-791, Republic of Korea;RIKEN, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan;RIKEN, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan
Venue:
Applied Soft Computing
Year:
2013

Citing 5
Cited 0

Simulated annealing and Boltzmann machines: a stochastic approach to combinatorial optimization and neural computing

Simulated annealing and Boltzmann machines: a stochastic approach to combinatorial optimization and neural computing
Neural computing: theory and practice

Neural computing: theory and practice
Natural computation and non-Turing models of computation

Theoretical Computer Science - Super-recursive algorithms and hypercomputation
Amoeba-based neurocomputing with chaotic dynamics

Communications of the ACM - ACM's plan to go online first
Simulation of Neurocomputing Based on Photophobic Reactions of Euglena: Toward Microbe---Based Neural Network Computing

UC '09 Proceedings of the 8th International Conference on Unconventional Computation

Quantified Score

Hi-index	0.00

Visualization

Abstract

We report on neurocomputing performed with real Euglena cells confined in micro-aquariums, on which two-dimensional optical feedback is applied using the Hopfield-Tank algorithm. Trace momentum, an index of swimming activity of Euglena cells, is used as the input/output signal for neurons in the neurocomputation. Feedback as blue-light illumination results in temporal changes in trace momentum according to the photophobic reactions of Euglena. Combinatorial optimization for a four-city traveling salesman problem is achieved with a high occupation ratio of the best solutions. Two characteristics of Euglena-based neurocomputing desirable for combinatorial optimization are elucidated: (1) attaining one of the best solutions to the problem, and (2) searching for a number of solutions via dynamic transition between the best solutions. Mechanisms responsible for the two characteristics are analyzed in terms of network energy, photoreaction ratio, and dynamics/statistics of Euglena movements. The spontaneous fluctuation in input/output signals and reduction in photoreaction ratio were found to be key factors in producing characteristic (1), while the photo-insensitive Euglena cells or the accidental evacuation of cells from non-illuminated areas causes characteristic (2). Furthermore, we show that the photophobic reactions of Euglena involves various survival strategies such as adaptation to blue-light or awakening from dormancy, which can extend the performance of Euglena-based neurocomputing toward deadlock avoidance or program-less adaptation. Finally, two approaches for achieving a high-speed Euglena-inspired Si-based computation are described.