%0 Journal Article
%T CUDAICA: GPU Optimization of Infomax-ICA EEG Analysis
%A Federico Raimondo
%A Juan E. Kamienkowski
%A Mariano Sigman
%A Diego Fernandez Slezak
%J Computational Intelligence and Neuroscience
%D 2012
%I Hindawi Publishing Corporation
%R 10.1155/2012/206972
%X In recent years, Independent Component Analysis (ICA) has become a standard to identify relevant dimensions of the data in neuroscience. ICA is a very reliable method to analyze data but it is, computationally, very costly. The use of ICA for online analysis of the data, used in brain computing interfaces, results are almost completely prohibitive. We show an increase with almost no cost (a rapid video card) of speed of ICA by about 25 fold. The EEG data, which is a repetition of many independent signals in multiple channels, is very suitable for processing using the vector processors included in the graphical units. We profiled the implementation of this algorithm and detected two main types of operations responsible of the processing bottleneck and taking almost 80% of computing time: vector-matrix and matrix-matrix multiplications. By replacing function calls to basic linear algebra functions to the standard CUBLAS routines provided by GPU manufacturers, it does not increase performance due to CUDA kernel launch overhead. Instead, we developed a GPU-based solution that, comparing with the original BLAS and CUBLAS versions, obtains a 25x increase of performance for the ICA calculation. 1. Introduction Analysis of brain imaging data has two intrinsic difficulties: dealing with high volumes of data (and often high dimensional) and a usually low signal-to-noise ratio due to persistent artifacts. A significant number of methods have been developed, usually based on some form of dimensionality reduction of data, to cope with these difficulties. Multivariate statistical analysis for the separation of signals is a widely studied topic of great complexity because of the large number of sources and the low signal-to-noise ratio, inherent in this kind of signals. Specific approaches have been developed to separate the signals generated by the study of those sources that contribute only noise, such as principal component analysis (PCA) [1], factor analysis [2], and projection pursuit [3], among others. Independent Component Analysis (ICA) [4¨C6] is one of the most effective methods for source separation and removal of noise and artifacts. The most emblematic example was the separation of audio sources in noisy environments [5]. In recent years, it has become a standard in brain imaging-electroencephalogram (EEG) [7¨C9], magnetoencephalogram (MEG) [10] and functional magnetic resonance imaging (fMRI) [11¨C13]. It has been used for the removal of artifacts arising from eye movements [14], but also for the selection of relevant dimensions of the data [15, 16]. In
%U http://www.hindawi.com/journals/cin/2012/206972/