Proposes new kernels for better classification performance.
problem Improving classification performance in computer vision.
method Developed GInt and NGMM kernels, and validated their effectiveness empirically.
result GInt kernel performs well in classification tasks, NGMM kernel outperforms GInt.