search

accosmin / nano

C++ library [machine learning & numerical optimization] - superseeded by libnano
MIT License
1 stars 0 forks source link

Hard coded convolution & correlation methods after the kernel size #9

Closed accosmin closed 9 years ago

accosmin commented 9 years ago

Implement and check if a custom 3x3, 5x5 ... kernel implementation is faster than conv2d_dyn or corr2d_dyn.

accosmin commented 9 years ago

These convolutions are not faster in the end than conv2d_dyn, which is doing more or less something similar.