Unexpected convolution output size

[Old posts from the commercial version of ArrayFire] Discussion of ArrayFire using CUDA or OpenCL.

Moderator: pavanky

Unexpected convolution output size

Postby rm9 » Thu Jan 30, 2014 5:59 am

Hi,

I'm doing the following convolution with expand = false:
Code: Select all
data(span, span, k) dims=[256 64]
W(span, span, j) dims=[8 64]
z dims=[256 64]


The input matrices are data and W and the output is z.

I expected the output to be in size [249 1] since I assume expand=false means no padding of zeros.
Is that wrong? How can I achieve a convolution like MATLAB's 'valid' option?

Thanks.
rm9
 
Posts: 54
Joined: Thu Jan 30, 2014 5:44 am

Re: Unexpected convolution output size

Postby rm9 » Thu Jan 30, 2014 11:53 am

Sorry for the quick posting, I searched the forum and understood that it is not a 'valid' convolution as I expected.
rm9
 
Posts: 54
Joined: Thu Jan 30, 2014 5:44 am

Re: Unexpected convolution output size

Postby shehzan » Thu Jan 30, 2014 11:55 am

Please see this page and let me know if it help: http://www.accelereyes.com/arrayfire/c/ ... 47f35bfd40
----
Shehzan
Developer
AccelerEyes
User avatar
shehzan
 
Posts: 121
Joined: Tue Feb 12, 2013 7:20 pm


Return to [archive-commercial] Programming & Development with ArrayFire

cron