Extracting Fuzzy Rules to Compare Genetic Algorithm-Generated Motoneuron Models

Date

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Spinal motoneurons that have been active for prolonged periods of time exhibit different electrical properties than their less active counterparts, suggesting that prolonged neuronal activity may change how electrical signals are transmitted through the neuron. Understanding how these spinal motoneurons integrate their input signals and modulate their output is important, with implications for rehabilitation, advanced prosthetics, brain-machine interfaces, humanoid robotics, and other biologically-inspired systems. To investigate what changes may take place within a spinal motoneuron following prolonged activity, a genetic algorithm was employed to generate two distinct groups of spinal motoneuron computational models. The first group (control) simulated less active neurons while the second group simulated neurons treated with high K+, which mimics persistent activation. The models had nine variable parameters, each a conductance related to a specific ion channel present in the motoneuron. To evaluate fitness for each computational model, fuzzy logic was used to assign membership in fuzzy sets corresponding to two separate objectives: current threshold and input resistance. To mine rules from the generated data, correlations were looked at between each fuzzy set and each parameter. While no rules were successfully mined in this research, some interesting results were produced. Some relationships that exist between parameters within the control (less active) models, do not seem to exist in the treated models. Relationships were also found between parameters that exist in both groups of models, suggesting a possible co-regulation of the genes which express those traits.

Description

Keywords

Citation

DOI

Collections