Sunday, December 24, 2006

New music

Added to my iTunes/iPod archive since October 15th:





NIPS 2006 Whistler group photo


At the top of Whistler Peak during the ski break in the NIPS workshops. Left-to-right: Rick Zemel, Sham Kakade, David McAllester, me, Sam Roweis. Black Tusk in the background.

Saturday, December 23, 2006

Belated ski trip report


After NIPS in Vancouver and Whistler, I stayed to ski three days. On the last day, I took a backcountry tour guided by Conny Amelunxen, a cool, informative guide who mentioned modestly some climbing and guiding in South America, which turns out to be rather extensive (just search for his name).


We left from the bottom of the Harmony chair as the last snow showers behind the previous day's storm cleared out. We skinned up Burnt Stew Trail and dropped onto the Flute Creek drainage. After some beacon practice, we skinned up to a pass, skied down on a mix of powder and wind-affected stuff, had lunch, skinned back up, and dropped for the best run of the day onto the Oboe Creek drainage. Excellent powder was had all the way to the bottom (see the picture). The skin out was long. Late in the afternoon we came back to another pass back into the Flute Creek area, dropped into wind scoured and then a thin and breakable wind slab and finally a few turns on deep wind deposited powder down to the bottom of the drainage, for a skin up to Burnt Stew, and a twilight schuss back to the village.


We were lucky with the weather and with the conditions, which were safer than one might expect after a big storm.

Catching up with research reading (1)

The Dichotomy of Conjunctive Queries on Probabilistic Structures. [cs.DB/0612102]:We show that for every conjunctive query, the complexity of evaluating it on a probabilistic database is either PTIME or #P-complete, and we give an algorithm for deciding whether a given conjunctive query is PTIME o #P-complete. The dichotomy property is a fundamental result on query evaluation on probabilistic databases and it gives a complete classification of the complexity of conjunctive queries.

The question about this very interesting line of work: what price do we pay for probabilistic inference, as opposed a weaker semiring-based scoring?

This is also a test of posting from NetNewsWire via ecto.

Tuesday, December 5, 2006

Testing ecto as a client

As I noted before, MarsEdit doesn't work well with Blogger Beta. According to NewsGator, they'll be solving the problem when the new Blogger API stabilizes. In the meanwhile, I decided to try out ecto, which has a patched version that is claimed to work with Blogger Beta. This posting is a test.

Saturday, December 2, 2006

MarsEdit and Blogger

The interaction between MarsEdit and Blogger is rough. MarsEdit doesn't seem to know about titles for Blogger posts, although it handles titles for Radio posts correctly. This has been reported on the MarsEdit support forum. I had several MarsEdit crashes and no connection to this blog until I switched from "Blogger" to "Other Blogger-compatible." Clearly, this needs work.

Do animals tell lies?

Do animals tell lies?: (we have ferrets, so we are biased) [...] Berkeley and Stanford are notable (the title comes from an animal behavior course) (Via tingilinde.)


I've been reading Franz De Waal's Primates and Philosophers. De Waal is clear that the answer is an emphatic "yes" from the evidence he has gathered from is extensive studies of primates. Besides this direct evidence, the indirect evolutionary evidence he presents is quite convincing to me. De Waal is a wonderful writer. Highly recommended, with discussions by Robert Wright, Christine Korsgaard, Philip Kitcher, and Peter Singer as a bonus.

Trial move to a new home

I've been blogging at Fresh Tracks, hosted by Userland, for several years. It served me well, but the Radio Userland application has not kept up with changes in OS X and the Mac platform, so I'll be trial-blogging here. I hope it works out.