We just posted the list of accepted papers to second workshop on Large-Scale Recommender Systems and the Netflix Prize Competition. Here are the titles:

• Jinlong Wu and Tiejun Li. A Modified Fuzzy C-Means Algorithm For Collaborative Filtering
• Gavin Potter. Putting the collaborator back into collaborative filtering
• Andreas Toescher, Michael Jahrer and Robert Legenstein. Improved Neighborhood-Based Algorithms for Large-Scale Recommender Systems
• Tamas Kiss, Miklos Kurucz, István Nagy and Andras A. Benczur. Large-scale recommenders based on Association Rule Mining
• Oscar Celma and Pedro Cano. From hits to niches? or how popular artists can bias music recommendations
• Domonkos Tikk, Gabor Takacs, Istvan Pilaszy and Bottyan Nemeth. Investigation of Various Matrix Factorization Methods for Large Recommender Systems

It is a lot of work to grind through a research project and get an interesting paper out of it. Mostly, you have to be patient enough and work everyday at it. If you follow a sane process, it is difficult to fail entirely.

Picking the right research question is very important however: it is difficult to recover from a bad choice of topic. There are at least 3 types of good research questions: 1) explain with a theoretical model a (puzzling) experimental observation 2) improve by at least an order of magnitude an existing technique 3) make up a new problem and be the first to propose a solution (I call it Turney’s way).

I now believe that options 1 and 3 are far better than option 2. To illustrate my opinion, here is a little scenario:

• read a paper;
• think to yourself: I could improve this idea ten times over;
• get excited, dream of fame, start crafting a paper;
• late on Friday night, realize your contribution is tiny;
• keep going (because you have invested so much);
• months later, publish a weak paper.

So I submit to you Lemire’s first rule of good research: you must either be trying to explain puzzling experimental results, or be inventing new problems. In some sense, it amounts to discarding the “engineering way” which is to constantly perfect existing techniques.

Further reader: I have written much about how I think one can write a good paper and about my usual research process.

Many of my colleagues never mark assignments. I tend to mark papers on nearly a weekly basis. Why am I doing this? Because I believe that marking assignments is the best way to identify the weaknesses in my courses and learn from my students.

Many researchers never implement their ideas. They let their students do the lowly implementation work. I almost always do at least some of the implementation in all projects I work on. Why am I doing this? Because I believe that you never really understand an idea, even your own, until you have put it in practice. You never know how it feels to ride a bicycle until you have done it once, no matter how great your mind is.

On an unrelated note, my friend Yuhong came over during the week-end. She is a brand-new Software Engineering professor at Concordia University. She bought my wife some gorgeous flowers. Nice.

I just read a great essay by William Deresiewicz, an associate professor of English at Yale. His message is clear: heavy-league education is flawed.

Here is the killer sentence:

It’s no coincidence that our current president [Bush], the apotheosis of entitled mediocrity, went to Yale.

Via Sébastien Paquet.

See also my posts It may not matter all that much where you go to college and The 2 myths getting students into heavy-league schools.

Disclaimer. I am a University of Toronto graduate. The closest thing Canada has to an Elite education, I would guess.

I suffer from exhaustion. In the last few weeks, I had to resign from a few hats I wore. I resigned as union treasurer and I resigned as chair of the IT M.Sc. degree. Today, I stood up my friend Yuhong on a lunch date. Too much to do, too little time. I have reached a breaking point.

I have always hated meetings. I prefer to work alone at my desk, with the occasional email. I realized recently that blogging makes meetings feel even worse.

There are many types of information you will not get through traditional channels. Peter Turney’s latest post is one such example. He basically says that simplicity is just one type of bias: the simpler solution is not necessary better. Wow.

I am sure there are many people who are within meters of Peter right now, and they missed his post. They are probably busy preparing some management meeting. What a waste of time!

If you have time for meetings, you do not spent enough time gossiping in the blogosphere.

Do we need meetings? I just want to be left alone to reflect, write and read. If more of us did this, I am sure humanity would be collectively smarter.

• I have not used Microsoft Office in over 5 years. I use Mac OS and Linux.
• I never use my employer’s email service. Prior to Google Mail, I used a private provider and forwarded my work email there.
• I have never driven to work, in the last 4 years.
• As a researcher, I do not belong to any one community.
• I keep teaching an university-level XML course, even though I have been ridiculed for teaching such lowly technical issues.

This is how most people understand purity in Science:

As for myself, I measure purity on a bandwidth scale: the more feedback the researchers get, the less pure they are. I should maybe use another term.

(Thanks to Steven for pointing this comic to me.)

Sufficient focus is necessary to be smart. The corollary is that distractions may turn your brain into mulch. There several conditions to sufficient focus:

• a sense of urgency: without a strong need to get the task done, long term focus is difficult;
• the dismissal of external stimuli: either you make sure not to be disturbed, or you can filter out the distractions;
• mental readiness: sometimes your mind will simply not focus before you rest.

Tag clouds are an interesting visualization technique because, unlike bar charts, you can easily display 30 or 50 weights in a compact figure. Moreover, because they are a 2D structure, you can more easily cluster similar tags together. The Tag-Cloud Drawing problem is the optimization of the layout of the tag clouds. It is somewhat related to the Graph Drawing problem.

Recently, Fujimura et al. showed how to scale tag clouds further… up to 5,000 attributes!

We use a topographical image that helps users to grasp the relationship among tags intuitively as a background to the tag clouds. We apply this interface to a blog navigation system and show that the proposed method enables users to find the desired tags easily even if the tag clouds are very large, 5,000 and above tags. Our approach is also effective for understanding the overall structure of a large amount of tagged documents.

I really think that tag-cloud drawing is a topic deserving of more attention. It is both a fun and practical problem.