Title:Neural Network Modeling and Analysis of Turn Duration Time Changing of Silkmoth Using Genetic Algorithm
Volume: 2
Issue: 2
Author(s): Ryosuke Chiba, Sunao Hashimoto, Tomoki Kazawa, Ryohei Kanzaki and Jun Ota
Affiliation:
Keywords:
Genetic algorithm, mobiligence, neural network modeling, silkmoth zig-zag walking.
Abstract: In this study, we investigate the reproductive behavior of the male silkmoth, Bombyx mori.
When a male silkmoth senses the sexual pheromones of a female through its antennas, it shows a certain
walking pattern by which it approaches the female. Interestingly, the degree of pheromone stimulation
influences the turn duration time in this pattern. This walking pattern is considered to be generated
in the silkmoth brain, specifically in the lateral accessory lobe (LAL) and the ventral protocerebrum
(VPC) domain, which control physical movements However, the system responsible for this
behavior remains unknown. In this study, we investigate the generation of this behavior through a neural network model
of the LAL and VPC domains. Specifically, we model many neurons in the silkmoth brain using one artificial neuron and
estimate the strength of each connection using a genetic algorithm between 10 neurons that represent neuron groups with
a fitness function of turn duration time. The model of the silkmoth brain is verified and evaluated from both engineering
and biological viewpoints. The modeling results show that only 6 LAL- VPC regions can make the turn duration time
shortening and buffering regions play very important roles in the reproductive behavior. Subsequently, we developed a
new hypothesis that a male silkmoth adjusts its walking pattern using the time delay of transition the signals between lateral
and bi-lateral regions.
