Thursday, August 21, 2008

SciPy conference

I'm at the SciPy conference today. I didn't sign up, but Alex gave a keynote and I got to ride on his coattails. (Although I'm being hit up for presenting in the future...)

Alex talked about the problems of black boxes and that "abstractions leak". He gave some nice examples of why you want to understand at least a couple of levels lower than the abstraction you're using (eg., statistical packages won't fix sample bias problems).

The next speaker after break talked about graphing networks - a package called NetworkX. This was a really gorgeous graphing package that looked like it has a really simple, usable API. I particularly liked that in the examples, they used
import networkX as NX
rather than the usual "from blah import *". It's important, particularly in tutorials and FAQs, imho, to SHOW where all the functions you're using live. Explicitly showing all the dots makes it clear where these things live. Using import * hides that. My compliments to the NetworkX people for not hiding things.

One of the use-cases they mentioned at the beginning was for synchronizing oscillations, and I immediately thought of neo-cortical neural synchronization, which they mentioned at the end when they demo'd their oscillation synchronizer.

NetworkX has a very pretty gallery of graphs to show some of its possibilities. Check it out if you have any graphing or complex network modeling needs.

The next talk was on intervals. It harked back to Alex's talk, which was amusing. Pay attention to your numbers - including the algorithms and packages you're using to calculate things. It was fun to watch - it reminded me how much I actually *like* math. (I was the geek who used to solve quadratic equations for fun.) I'm not so good with trig and logs (although I've gotten better). Anyways - I really enjoyed the intervals talk - it was fun to see a purely mathematical coding talk; it's pretty rare.

We're going to have to head out this afternoon to the airport. If I catch any more presentations, I'll post about them.

Social-wise, the SciPy people were very nice. We got picked up at the airport, driven to our hotel, and they took us to dinner. This morning, breakfast included *protein* (Yay!) not just carbs. And I got to meet another Minnesota girl - who's now working in neuroscience down at Berkeley. We're going to get together when we're back.

In short - I've really enjoyed being here and I plan to attend SciPy in the future. If you are into scientific uses of Python at all - I encourage you to attend.

1 comment:

bit twiddler said...

import networkX as NX
not the usual "from blah import *"

Code with clarity,
not obscurity,
and shed light,
not shadows.

Thanks for posting your tip!