AMD64 version?
Hello,
I've been testing out various recommendation algorithms and I came upon suggest. I have a data set of around 200,000 users and 800,000 items with around 16 million ratings from 1 to 5. I'd like to do some more testing with suggest, but is there any way to get a library compiled for an AMD64 system or the library source?
Thanks in advance!
Submitted by max on Tue, 2007-04-17 05:26
»
- Login to post comments
RE: Yes we too would be very
Yes we too would be very interested in a 64 bit version of the binary.
RE: Also interested on AMD64 version
I'm also interested on the AMD64 binary library (I intend on releasing a python wrapper for libsuggest under the GPL).
The source code would also be of great value. As part of my Msc. thesis I'm investigating the usefulness of integrating visual image features into image recommenders for social networks like Flickr, Zooomer etc. Some early work is available at (http://fs.imgseek.net/).
Thanks in advance and best regards!
RE: This may take some time, as
This may take some time, as it will require some non-trivial changes.
RE: Well, the end goal of my
Well, the end goal of my project is a recommendation system for a for-profit site (it's free for anyone to use but profits off of ad revenue). I know you'd rather not release the source for any business purposes, which is completely understandable. That being said, I'd rather you didn't do the work solely for me in that case.
I've been working on a system that manages to do rating prediction with an RMSE of around 0.9-1.1, but due to the way it was designed, doing top-N suggestions is cumbersome and slow. Once finished, my plan is to release the code under the GPL, as from spending a lot of time on this project there are a plethora of white papers on the subject, but short of your program almost everything that's been made is proprietary. Luckily due to the recent netflix challenge, lots of good information and small code samples have been popping up, but no one has released a working system that's anywhere near complete.
I was hoping to get an idea of how you managed to pull off top-N, so quickly, but in reality I should just overcome my own design issues.
If in the future you wish to release the source for the library, I'd be completely willing and able to help port it over to a 64 bit architecture.
Cheers!