good-bye money

I ordered two math books from amazon so that I don’t have to enter into a recall war at the library with the other person who seems to want them so badly. I’m really digging the one I’m reading now though — Differential Geometry and Statistics. It’s got a gentle introduction to geometric concepts when dealing with things like fitting statistical models to data and will hopefully will give me a more sophisticated view on modeling. It’s way easier to read than Amari’s monograph on the same subject, which has a whole chapter devoted to a complex theoretical apparatus with no examples. Saying things like “the curvature looks like this function, which we call the Fisher information” is not nearly as helpful as “remember the Fisher information? It looks like this. This turns out to be a special case of the a general concept called the curvature.”

I need to read the material bottom-up. Or sideways, depending on where you think Statistics is relative to Differential Geometry. Actually, since it’s geometry, we’re trying to free ourselves from the tyranny of coordinates, so “up” and “sideways” are pretty meaningless altogether.

public defenders

An article at Left2Right reminded me of the movie Slam which I saw recently. Compared to the defenders discussed in the article, that one seemed competent and honest, if a bit hopeless and mean. Essentially he told the protagonist to plea bargain or else, which is a tactic, not an argument. But the article is right — don’t blame the PDs, but blame the underfunded system that drives people away and encourages grifters.

yawn

In case the old jive filter still amuses you, someone has written Gizoogle. According to its version of my homepage:

This was one of tha first projects I worked on at Berkeley spittin’ that real shit. We present a way of sippin’ a priority queue fo` optical data, where pimpin’ is not feasible.

Actually, I don’t find this sort of thing that funny after about 30 seconds.

too many problems

… and not enough solutions:

  • Matching source and channel models for large sensor networks : I want to characterize when a fully-distributed set of gains can “match” a linear/matrix observation model with a linear/matrix communication channel model. I think it’s a broad class, especially as the number of gains (which is the number of sensors) increases.
  • Dense sensor network scaling laws: under some better constraints, it should be possible to show that the total capacity does not scale with the network size for dense networks.
  • Causal jamming relays: a new take on relays as potential jammers. You can’t trust anyone.
  • Fast distributed consensus: not quite sure where this is going, but it has to do with creating a speedy algorithm for calculating the average of a function on a graph.

There are, of course, other problems, but unless I write these down I’ll be liable to forget them when the next shiny object comes along.

maybe we can get along

I recently installed Ubuntu on the computer I use at work, and am so much happier than I was running Windows. Ubuntu is a more user-friendly version of Debian, which is a godsend for laptops, since the installation process does all the hardware detection for you. I had tried Debian and getting X to work was a real pain. I then switched to Fedora Core 3 and decided it made me feel like I was in a straitjacket. Ubuntu is “just right” for me I think. Also, the releases have amusing names, like “Warty Warthog” and “Hoary Hedgehog.”

While browsing for extra LaTeX plugins, I came across the following:

arabtex (3.11-5) [universe] — Arabic/Hebrew macros for TeX/LaTeX

Arabic and Hebrew, eh? Maybe we can all get along.

LaTeX templates for conference posters

I hacked together a LaTeX template for conference posters for the Wireless Foundations Center. After searching around on the web I couldn’t find a package that was (a) easy to use, (b) allowed for arbitrary layout, and (c) supported many different poster sizes. So I hacked together 2-3 other style file ideas from other people and made this template and associated style file. I’ll try and add features over the summer, but that’s a really low priority right now.

I’m using the package to make a poster for tomorrow’s day conference at Stanford and it’s going pretty smoothly, so I think it’s reasonably usable.

stealth soloist selection

Ari had a brilliant idea for a way to run chorus rehearsals. Soloist selection will be at random for each piece, but you won’t know if you have a solo until it’s time to sing it.

The corresponding conducting technique is to give the upbeat and then on the downbeat suddenly point at the soloist-to-be. This technique is adapted from a certain conductors-who-will-remain-unnamed’s habit of switching soloists at the last minute. The advantage is that everyone will be looking up out of their music, so it’s not all a wash.

tenors

I promised a while back that I would talk about tenors: Peter Pears, Fritz Wunderlich, and Ian Bostridge.

Manu lent me a cd of Ian Bostridge singing Schumann’s Dichterliebe with the (to me at the time) ludicrous assertion that it was significantly better than Wunderlich’s approach to the same song cycle. After long deliberation between the two recordings, I would agree that they are significantly different but the absolute quality depends on what you are in the mood for.

Wunderlich is a master of technique to the point that even the heavy Romanticism of Schumann seems somehow classical. He gives free reign to his vibrato in a natural way, it seems. Barring that effect, the songs are simple yet beautiful. He lets the sonic texture of the music speak for itself. Bostridge, on the other hand, takes a much more active role in the interpretation of the music. At times it sounds too meddlesome — in the song “Ich hab’ in traum geweinet,” he sounds too overwrought to me for such a spare and clinical accompaniment. On the other hand, in songs like “Ein Juengling liebt ein Maedchen,” Wunderlich sounds lumbering compared to the nimble and light Bostridge. It’s a toss-up for me, I think. I’m not sure I’ll ever be able to implement my own interpretation of the Dichterliebe, but it lies somewhere between then. But having listened only to the Wunderlich, Bostridge’s interpretation is a breath of fresh air.

Bostridge’s vocal style leads me to talk about Peter Pears, who was Britten’s lover and partner (had they been in MA I’m sure they would have gotten married). Britten wrote many pieces for Pears that highlight what some characterize as an unusual voice. I don’t think it’s unusual at all, but rather a more natural voice than we hear from other classically trained singers. Pears sounds like he comes from the same “warbler” tradition that appears in the flashback scenes of The Singing Detective. He is full of vibrato and perhaps sounds ponderous in his low range, but his tone is agile, for lack of a better word. He glides effortlessly through Britten’s Nocturne, for example, a feat which most other tenors would find difficult to be sure.

All of these singers are worth a listen. It adds a whole different dimension to listen to such wildly varying interpretations as Wunderlich and Bostridge, and it’s worth hearing Britten as sung by his intended tenor to get the feel of his solo voice music.