Archived

This topic is now archived and is closed to further replies.

as31415rin

performance of NN simulators ?

Recommended Posts

I just havn''t found results for the performance of NN simulators on normal PCs I wrote a simulator and wanted to know how fast it is. On a iCel466 a 16-16-16 feedforward net using vanilla backpropagation and log (tanh) activation function i get 16MCPS+5MCUPS (11MCPS+4MCUPS) @$3.1415rin

Share this post


Link to post
Share on other sites
I am interested in neural networks but havent found any sources on how to simulate them, could you give me some examples, possibly for vc++? Well anyways i''ve never once read on testing the performance of a net but know a lot about them. I am a vc++ newbie, but like to study them and come up with cool models i would like to simulate, any help would be much appreciated. Sorry i couldn''t help you with your question.
Jonathan Redig
redigjohn@yahoo.com

http://www.GamerzParadise.com
http://www.cheapefilms.com

Share this post


Link to post
Share on other sites
Others have measured their neural networks this way (see, for instance, California Scientific Software, at http://www.brainmaker.com/). I will suggest that this is a rather limited measure of neural network performance, though. Other model qualities are of greater interest, such as model accuracy and ability to learn incrementally.



quote:
Original post by as31415rin
I just havn''t found results for the performance of NN simulators on normal PCs

I wrote a simulator and wanted to know how fast it is. On a iCel466 a 16-16-16 feedforward net using vanilla backpropagation and log (tanh) activation function i get 16MCPS+5MCUPS (11MCPS+4MCUPS)

@$3.1415rin


Share this post


Link to post
Share on other sites