[mlpack] GSOC 2016 "Alternatives to neighbour-based collaborative filtering"

Ryan Curtin ryan at ratml.org
Mon Mar 14 09:52:00 EDT 2016


On Sun, Mar 13, 2016 at 05:38:22PM +0800, Zhaoduo WEN wrote:
> Hi, Ryan,
> 
> I just knew about GSOC today and I hope it is not late to introduce myself.
> 
> I am Zhaoduo Wen, a senior student from Beijing University of Posts and
> Telecommunications.I've been working on data mining and machine learning
> problems for a year and had some experience on recommender systems.
> 
> >From what I know, although collaborative filtering is faster in prediction
> than matrix factorization framework, either itemKNN collaborative filtering
> or userKNN collaborative filtering has some drawbacks. One major
> disadvantage is that they suffer from low accuracy since there is
> essentially no knowledge learned about item characteristics so as to
> produce accurate recommendations. However, linear sparse model performs
> better both in prediction accuracy and running time. I have read related
> papers and I was lucky to listen to the author's presentation.
> Consequently, I prefer to using a sparse linear model as the alternatives
> to neighborhood-based collaborative filtering.
> 
> I am enthusiastic for contributing to this project as I will be extremely
> excited if I finish this project and someone uses it in future. I once used
> a library for large linear classification (LIBLINEAR), which has a high
> citation times on google scholar. I was impressed by its fast and accurate
> performance. I wish I could write one someday. I believe GSOC would be a
> good beginning.
> 
> What is your opinion about my proposal? Hope to receive your reply. Thanks.

Hi Zhaoduo,

Can you provide a link to the paper that you are proposing to implement?

Also, it would be a good idea to take a look through the existing CF
code to see how the sparse linear model you are proposing would fit into
the API.

Thanks,

Ryan

-- 
Ryan Curtin    | "Leave the gun.  Take the cannoli."
ryan at ratml.org |   - Clemenza


More information about the mlpack mailing list