Summary: This commit uses "math::Transpose" instead of "kernels::Transpose" for more possible optimized routines in the future.
