Whoo, got an google appengine test account. Now what'll I do with it?
@mikiobraun Twitter Memorial
19,827 tweets · 2008–2024 · 1046 threads
2008
Coming home to Germany, jetlag is almost worse than when going to Canada this time... . How many days till Christmas?
NIPS is finally over. I guess our banquet talk went over well. Only people now really seem to want even more T-shirts from us ;)
... unless one of the kernels is really bad.
It seems averages of kernels work slightly better than MKL on his bioinformatics application. And it's much faster too :)
Kernel Learning WS: W.S.Noble compares MKL to unweighted averages of kernels.
Looks like luckily, the -12°C is not going to happen today.
Next we will have a short presentation on 2 projects related to databases which leads up to the discussion on reproducibility and datasets
Next talk: Brian Tanner on RL Glue http://mloss.org/software/view/151/
Next talk: Troy Reader on Model Monitor, a tool for evaluating classifiers under distribution shift http://mloss.org/software/view/137/
Jeremy Hill on BCPy2000, a python framework for BCI http://mloss.org/software/view/145/
Coffee break, will continue at 5:20.
Last talk before the coffee break: Joris Mooij on libDAI, a library for graphical models (5:00-5:10)
Next talk at about 4:40: Nicolas Usunier on Nieme, a toolbox focusing on energy-based methods http://mloss.org/software/view/49/
Next up: Ville Tuulos on Disco, a distributed computing framework for python http://mloss.org/software/view/140/
Try "import antigravity" in python 3.0
Afternoon session: First talk John Hunter on matplotlib, a plotting library for python http://matplotlib.sourceforge.net/
Next up before the break: Discussion on how to improve interoperability.
Next up: mdp (modular processing framework) by Tiziano Zito http://mloss.org/software/view/60/
Talk by Davide Albanese: mlpy, a toolbox for python (10:00-10:20) http://mloss.org/software/view/66/
Current Schedule Adjustment: unfortunately, +30min... . Two more talks before the discussion: mlpy, mdp two toolboxes in python
Talk by Alexandros Karatzoglu on kernlab, a ML toolbox for R http://mloss.org/software/view/45/ (9:40-10:00)
Next up: unplanned coffee break. Program will continue at 9:30-9:35
Tobias Glasmacher's talk starts now (9:05-9:25)
In the end, Ronan discloses that also his slides were actually rendered by Torch.
Schedule time adjustment: Currently about +13 minutes
Next talk at about 9:03: Tobias Glasmachers on Shark http://mloss.org/software/view/70/
Ronan's talk starts now (8:38-8:58)
Next talks will be 20 minutes each, so be sure to be here in time if you're interested... .
Next up at about 8:30: Ronan Collobert on Torch http://mloss.org/software/view/128/
"More ways to contribute beside code: web site work, mailing list maintenance, support, etc."
"... and of course domain specific packages.."
"... or if you're interested in just-In-time compilation..."
J. Eaton: "If you're interested in documentation or release management, talk to me"
Next talk (7:50-8:35): John Eaton on his matlab-clone octave http://is.gd/aCgN
Soeren Sonnenburg giving overview of MLOSS activities in the last two years.
MLOSS workshop about to start in Alpine CD at the Westin. First talk after the intro: John Eaton on Octave.
Main NIPS conference is over. Mini-symposia were definitely a win. Bus transfer to Whistler. Our workshop starts tomorrow at 7:30am.
Very entertaining talk by Steve Seitz showing how you can do 3d reconstruction based on photos from flickr http://is.gd/beeI
Talk in comp. photography mini-symposium makes Seinfeld jokes.
Sky also finally cleared up. First time to see the mountains across the bay. ^_^
Final day of the conference. Mini-symposia in the afternoon and then we'll go to Whistler by bus for the workshops.
Most method-centric (a.k.a. application-agnostic) session so far :)
Interesting talk by Tong Zhang on robust feature selection by combination of feature add and deletion steps.
Biggest event this morning: Wireless access broke. Internet still shaky. Session chair suggested, we should listen to talks instead ;)
Jetlag compensation hopefully completed and all mloss t-shirt handed out. Workshop banquet talk still missing...
From the overview paper (s.b.): "The key observation is that, in many applications, the uncertainty does not depend on the decisions".
Overview paper on online stochastic optimization by Pascal Van Hentenryck http://is.gd/aUoH
Invited talk on stochastic optimization. Motivation sounds quite interesting so far.
This year we again brought several boxes of mloss t-shirts. Make sure to grab your free t-shirt before they are gone!
Coffee break ;)
First session on neuroscience and brain computer interfaces.
Waiting for our turn at the buffet. Severly jetlagged.
Safely arrived at Vancouver. Lucked out on the in-flight entertainment: video on demand!
Amsterdam... .
Don't know why it's called "travelling". It should be called "waiting".
Replying to @mdreid
... and like my Ph.D. adviser used to say: Bayes Theory is the Statistical Physics of Machine Learning. :)
Replying to @mdreid
Thanks! My daughter is 20 months now, hope her cough from last week is now gone for good... .
Last day at the university before flying to Vancouver for the NIPS conference. Some last emails and things to pack...
By the way, I'll plan to do some live-blogging for our workshop so that people can tell how far were behind schedule ;)
Less than two weeks till NIPS and we're still not set on our Workshop banquet talk... . ;)
If you're interested in kernel methods, I've put up an overview about my latest JMLR paper. http://is.gd/7OHm
Finalizing the brochure for our MLOSS workshop. But I think that will be about it soon.
Finalized the program for our NIPS workshop on machine learning open source software. Wasn't easy, but I think we have a nice schedule.
We're working hard on finalizing the program for our NIPS workshop today. We have been waiting for some late reviews.
Living so far to the east of one's timezone can really suck in the wintertime.
very cool: http://www.xkcd.com/482/
I found that the closing chapters of the second book of Dune harmonizes very well with Snow Patrol...
THESEUS meeting finished ahead of time, ready to leave my office for the weekend :)
Done with lecture, now on to THESEUS meeting... .
Preparing slides for tomorrows lectures. I guess this will take all day long.
Replying to @paulbuenau
@oberwolfach sehr unterhaltsam, weiter so!
Back at Fraunhofer FIRST for our annual Berlin-Tuebingen seminar.
The disk died on our mloss.org servers, so we're extending the deadline to our workshop http://mloss.org/workshop/nips08/ till next monday
Guess its going to be rainy day here in Berlin...
Very nice, our NIPS workshop got accepted! http://mloss.org/workshop/nips08
Halfway through till the semester starts, and again writing project proposals... .
Hacking jblas (http://www.jblas.org). Finally some free time to do some major restructurings... .
Replying to @fluchtpunkt
: und, wieder gesund?
Der Zug hält wirklich an jeder Hütte. Jetzt weiß ich wenigstens wo Hildesheim ist.
Irgendwie steigt von meinem Büro aus Richtung Helmholtzstraße Rauch auf... .
Replying to @fluchtpunkt
hab' Dich mal zu meinen Followern dazugefügt...
Hanging out in Lübbenau on our group trip
Langsam weiß man, was man an Deutschland-Portugal hatte... gähn
Als wir landeten ging über dem Alex gerade ein Gewitter nieder, mit den entsprechenden Winden quer zu unserem Anflug...
Gut, ich hab's bis eine halbe Stunde vor Abflug sogar bis zum Gate geschafft. Aber langsam werde ich zu alt für sowas...
Hm. Eine S-Bahn ist ausgefallen. D.h. Ich hab' nachher 8 Minuten, um eine Bordkarte zu bekommen :(
Soweit so gut, nur am Flughafen ist es etwas eng kalkuliert, fürchte ich.
Jetzt weiß ich wenigstens, daß Tagungszentren nicht nur günstig sind, sondern daß man auch um 9 die Zimmer räumen muß.
Rheinländer = Gäste, nicht der Pfarrer, der war Religionslehrer.
Sektempfang. Sehr lustiger Pfarrer. Rheinländer sind auch einfach erfrischend.
Ah, des Rätsels Lösung, die Trauung beginnt erst um 3.
Hm. Kirche ist hier aber sonst kaum einer.
Nach 4 Jahren Berlin ist man die altstädtisch zusammengedrängten Menschenmassen gar nicht mehr gewohnt...
Sogar pünktlich angekommen... Weiß gar nicht ob ich schon mal hier war.
Noch einmal umsteigen, dann bin ich in gut 2h da.
Jetzt weiß ich auch, warum der "Bär" heißt, der Kerl ist gut 2m groß. Stelle gerade fest, daß ich erst noch zum Hbf muss...
Bin gelandet. Thomas D war auch dabei, steht gerade vor mir :)
Sieht aus als wäre der Bär von DSDS mit an Bord...