Course talk:CPSC522/Collaborative Filtering

Hi Tanuj,

A nice page. It is perfect if you can give some explanation for your pseudo code and do some proof reading.

Bests,

Yu Yan

Thanks Yu Yan. I've added comments in the code.

Hey Tanuj,

• At one point, you used $q_{i}$ and $p_{u}$ to describe the latent features, then changed it to $f_{i}$ and $g_{u}$.

• If the correct way is

$(\sum _{<u,i,r>\in D}(r^{'}+b_{i}+c_{u}+(\sum _{\overline {s}}f_{ik}g_{ku})-r)^{2})+(\lambda (b_{i}^{2}+c_{u}^{2}+\sum _{k}(f_{ik}^{2}+g_{ku}^{2})))$
Then where are the indices $i$ and $u$ for the second term coming from...? What is the outermost bracket supposed to mean?

• What are the differences now that you've made this change?

Nice page! Could you clearly summarize the difference between old version and new version (maybe you can rename them to be more clear)?

Hi Tanuj,

Nice to see your page. I think your page requires updating your experiment and evaluation. Looking forward to your final version.

Regards Arthur

I see. I guess you were fast by an hour. :)