At last I’ve got the chance to write about the MMDS conference.

It was an awesome workshop: if with WSDM 2008 I had lost the faith in science, in these days I regained a deep devotion.

The schedule was very dense, most of the talks were enlightening, and there were two coffee breaks a day with fresh (sweet!) fruit, bagels and muffins. Everything for 25 bucks! It’s so nice being a student.

Like in February with the 24-70 f2.8, the parks around Stanford gave me a great opportunity of trying out my shiny new 70-200 f2.8 IS. What a wonderful lens! It’s sharper than the 24-70, probably because of the stabilizer. The results were not so bad:

Thanks to my usual care, I also managed to drop it, resulting in a nice scratch close to the bayonet and one on the hood. Luckily it’s sturdy (and heavy) like a tank.

There were also a lot of nice people! We formed a small group with other students, we had lunches together, we even went together to the gay pride when the workshop ended. I hope to meet them again in the future.

And now the boring part. Most of the talks were interesting, some where **very** interesting or even enlightening. These are my favorites, I strongly advice to go read the slides:

- Christos Faloutsos (Carnegie Mellon University)

TUTORIAL: Graph mining: laws, generators and tools - James Demmel (University of California, Berkeley)

Avoiding communication in linear algebra algorithms - Ronald Coifman (Yale University)

Diffusion geometries and harmonic analysis on data sets - Nikhil Srivastava (Yale University)

Graph sparsification by effective resistances - Leonidas Guibas (Stanford University)

Detection of symmetries and repeated patterns in 3D point cloud data - Piotr Indyk (Massachusetts Institute of Technology)

Sparse recovery using sparse random matrices - Gunnar Carlsson (Stanford University)

TUTORIAL: Topology and data - Partha Niyogi (University of Chicago)

Manifold regularization and semi-supervised learning - Nir Ailon (Google Research, New York)

Efficient dimension reduction - Lek-Heng Lim (University of California, Berkeley)

Ranking via Hodge decompositions of graphs and skew-symmetric matrices

Regarding the ones that I didn’t find interesting, probably I just didn’t understand them. Many required a basic knowledge about statistics (regression, distributions, moments, machine learning) which I know nothing about. In my next life I’ll try to take statistics more seriously.

I was very surprised by applications of topology to data mining. They are still not very refined, but could lead to beautiful techniques. I was also surprised by the fact that many speakers came from mathematics departments: in Italy I couldn’t imagine an algebraic topologist even **knowing** about data mining. I am reconsidering applying for a PhD in Mathematics (but not in Italy…)

Luca F. pointed me out the website of MMDS 2006, and the titles of the talks seem as interesting. Probably the slides are worth a look (if I only had time… đŚ )