The optimization for Libsvm training in analysis of LIGO data

The gravity wave is a physical phenomenon beyond time and distance. It could lead to the distortion of time and space accompanied with the release of energy. The Laser Interferometer Gravitational-Wave Observatory (a.k.a LIGO) is a large-scale physics experiment aiming to directly detect gravitational waves. The data on which we study came from LIGO Science Run 6 (S6) with more than 1000 channels observed. Libsvm is used to train the experiment data to predict what are true signals of gravity wave. It is a time consuming task with original dimension. Making use of grid computing could lead to a faster solution and LDG provides us computing environment. The paper concentrates on some optimizations in order to reduce the complexity of data to speed up the training progress and get a better ROC curve which indicates a larger ligo-auc.