Enhanced Growing Neural Gas for Spike Sorting

Abstract

The report proposed a robust spike sorting program which designed to be low computational power required, high performance, and have the ability to be further developed for specialized hardware implementation and for online clustering. In the evaluation, this program showed a 100% accuracy on computer generated Gaussian distributed and low-noise non-Gaussian distributed clusters. Also, this spike sorting program outperformed SPC when clustering simulated neural spikes. Our work also shows the ability to performclustering for actual experimental data. The program is programmed using Python and based on the algorithm named Enhanced Growing Neural Gas algorithm which is one of the latest spike sorting algorithm. In this project, the ability of spike clustering of our work using different types of data is tested. Also, limitations and advantages of our work would also be discussed.