Grenoble: Break from machine learning

France is currently quite cold. But I understand that we don't have nearly as much snow as you have in Ontario. We don't seem to have nearly as much time these days as we once did for some reason and it also gets dark very early and (as I said) it's pretty cold so I think for all of those reasons in some combined way we don't have as many photos to post as we once did and the original purpose of this blog was really to be a "photo blog" as Tom once suggested so that's too bad! But, I'm sure the photos will start coming back when it gets warmer and the days get longer and we (maybe) start having some more time somehow. In fact, this (the photo taking) might happen soon because we have a couple of friends coming at the end of the month and we'll likely do some kind of trip with them maybe to the South in search of some warmer temperatures (I think it's maybe 15 degrees in Marseille today)..

Unfortunately I got the response last night that our paper was not accepted to the WWW conference in May so I won't be going to Rio de Janeiro which is lamentable but life goes on I suppose. It's good work though (I think) but not fully developed yet and I think as we work on it and improve some things and get more results we'll get it at another conference (although in all likelihood it won't be as cool as the WWW conference and won't have as great of a location as Rio). Work is fine but moving along slowly as is usual it seems for me. I'm doing some work on "multi-task learning" which builds on the stuff being explained in the Grenoble Blog Tutorial Sets (TM) but here we try to improve our ability to learn by exploiting similarities in tasks. So for example let's say you're trying to classify patents into related/non-related to your area of interest but you only have 100 training examples and someone else wants to classify scientific papers into related/not-related to your area of interest but only has little training data as well. Well there's likely some latent features that are similar in each of the tasks and so even though they're not the same task if we learn them at the same time we can take advantage of the similarities between them and the extra training data from the other task. It is hard stuff in my opinion.

We have been training hard for our upcoming marathon in Annecy (the marathon is in Annecy, we haven't been training in Annecy). The training program that we're following calls for a lot of long runs (at least longer than what we've been used to during the week) and you know maybe that's part of where some of the time that we've been missing has gone. This week has been a bit of a "rest week" however which is nice and the kilometreage has been lower and we don't have to run very far tomorrow (Sunday) which is normally the big long run of the week. Therefore, we might actually take advantage of living in the mountains in winter time (we do so all the time in the summer time through biking) by doing some snow shoeing in the mountains tomorrow. The weather is supposed to be sunny but cold so hopefully we will go through with the plan! Another thing we signed up for a couple of days ago is the Grenoble-Vizille half-marathon. It's 2 weeks before our Annecy race so we need to be careful about going too fast but it should be a good "warm up" for the marathon and it's a really cool route following the famous "Route Napoléon". It's also neat because our running club (ASPA Meylan) is paying for our entry fee! What a nice group of people.

Ok so there are no new pictures (yet) but here are some old ones that you might recognize (I can't remember if I've put them here or somewhere or there or wherever before) and that will maybe help remind us that spring/summer is just around the corner...