A GPU-accelerated spike classification system based on CUDA architecture

Abstract

The present invention provides a GPU-accelerated spike classification system based on CUDA architecture, including a streaming data input module, a filter module, a spike extraction module, a spike feature extraction module, a spike clustering module, and a spike incremental clustering module ; Firstly, the system of the present invention has good noise resistance, can improve the accuracy of classification, and improve the speed of classification; secondly, all modules of the system are realized based on the CUDA architecture, so that the running time is greatly shortened, and real-time classification can be achieved on high-dimensional channels , the modular design of the last system has good reusability and scalability, and the system greatly facilitates the research of brain neuroscience.

Description

technical field [0001] The invention relates to the technical field of machine learning, in particular to a GPU-accelerated spike classification system based on CUDA architecture. Background technique [0002] Spike classification is a class of techniques for analyzing electrophysiological data that groups spikes into clusters based on the similarity of their shape. In principle, each neuron tends to fire spikes of a particular shape, the resulting clusters correspond to the activity of different putative neurons, and the end result of spike classification is to determine which spike is related to which of these neurons Correspondingly, the CUDA architecture (Compute Unified Device Architecture) is a parallel computing architecture for devices such as GPU (Graphic Processing Unit) graphics processors. Programming interface APIs include CUDA C, C++, OpenCL, etc. The CUDA kernel program is called the kernel function, which is a parallel computing function running on the GPU....

AI Technical Summary

Problems solved by technology

The clustering of the klusta system uses the Masked EM algorithm, which is based on the Gaussian mixture model and has good anti-noise performance. Although it achieves a relatively high classification accuracy, the system does not perform GPU acceleration, and the system running time is unacceptable. In addition, the system cannot Support online data processing; the kilosort system uses a template matching algorithm and realizes acceleration on the GPU at the same time, which can realize real-time processing of spike signals, but due to the instability of spike signals, this method may be affected by noise signals ;The mountainsort system clustering adopts the ISO-SPLIT algorithm, which has higher calculation efficiency on channels with lower dimensions, but it still cannot achieve real-time processing on high-dimensional channels

Images