Analog feedback in Euglena-based neural network computing - Enhancing solution-search capability through reaction threshold diversity among cells

Abstract

Microbe-based neural network computing, where the reaction of microbial cells to external stimuli is incorporated in the function of virtual neurons, has high potential for developing soft computing based on the survival strategies of the microbe. To utilize reaction-threshold diversity among the cells, we examined analog feedback in Euglena-based neurocomputing by solving a simple combinatorial optimization problem. The analog feedback was performed by blue light illumination to Euglena cells, where the intensity of the blue light was controlled using the Hopfleld-Tank algorithm with a sigmoid function. The solution patterns obtained with analog feedback had greater variations than those with digital feedback, implying that the solution-search capability of Euglena-based neurocomputing is enhanced by analog feedback. Moreover, the solutions obtained with analog feedback comprised one stable core-motive selection and additional flexible selections, which are associated with hesitation shown by humans when faced with a frustrated task. The study shows that using analog feedback in Euglena-based neurocomputing is promising in terms of incorporating the diversity of photoreactions of Euglena cells to enhance the solution-search capability for combinatorial optimization problems and to utilize the adaptive reaction of Euglena cells.