Sunday, February 26, 2012

The weeks fly by

It's been an immensely busy week and by yesterday afternoon I was already feeling spent. The previous blog post that I put up garnered a lot of interest and I spent a good few hours replying to emails with advice and support - many thanks for the links, the tales of other people's similar experiences and the general positive notes.

I spent the evenings before Thursday this week getting a talk ready to give at Munich's monthly Nerd Nite, held in a bar and usually with an audience of at least 200 of the cities keenest nerds. My talk, on atmospheric optics, was one I'd given a good few times before, to both physics departments and to the general public, but normally the talk lasts around an hour. Trying to condense it down to 15 minutes was not easy and I fear that my enthusiasm mixed with my naturally fast talking pace made it all a bit rushed. Still, the audience seemed to enjoy the pictures at the very least and so it was a good experience all in all. The previous talks of the evening, on polyphasic sleep and on the neuroscience of dreams were also interesting, though being in German I didn't get more than three quarters of the content.

Speaking of which, my German is coming along painfully slowly. I've upped the lessons to three a week, one with a private teacher and two in a beginner's class at a language school. The language school has a bit of a strange format however, with a very relaxed 'drop in when you want' approach which means that the class, normally of six to eight students, is made of of some who know nothing at all, and some who have been studying for a few months, so we spend a lot of the time with the teacher asking us what we'd like to do. This isn't her fault, but a problem with the format. Still, I have another couple of months there, and it's cheap and for now any input is good. I will also be starting a language exchange next week as the aspect that I have to work on now is simply getting out there and speaking. I can go for an hour or so talking in German with my private teacher, but it's ponderously slow and so I'm very self-conscious about doing this with anyone but my teacher for now - this is a natural language-learning hurdle that I need to get over.

On the physics front two of the projects are making progress and we've come up with a new way to tackle a rather interesting problem, but I'll leave the details of this for any papers which may come out of it.

On the front of various other directions of study I've been looking a little into ICA (independent component analysis), a method within machine learning, which is really a fascinating area. The idea is pretty basic, though the implementation can be complex and subtle, and can be explained with a very simple example. Say you have two microphones in a room and you have two people speaking at the same time. Each microphone will record a slightly different linear combination of the two voices. The question is then how you can take the two recordings, and knowing nothing more than the fact that the two signals were in no way correlated, disentangle them. It turns out that there is an extremely clever way to do this, and basically you can disentangle any form of uncorrelated signal, be it a visual signal, an auditory signal, or the readings from a brain scan. The latter use is the one that I'm most interested in, but I wanted to play around with doing this for mixed audio signals and images first, and with a few lines of Mathematica it turns out to be very simple. If you are interested in knowing about how this works then I can either recommend having a read through this tutorial, or ask me if you'd like more details about what I've implemented in code.

Of the online undergraduate courses that I and a number of friends are taking, three have now started: Model Thinking, plus Computer Science 101 (building a search engine) and Computer Science 373 (programing a robotic car). These are all very nicely presented but so far CS101 looks to be rather basic unless you don't know any programing at all, in which case it will probably be a great introduction. Model Thinking is actually surprisingly insightful, and just with a few examples he has shown how powerful this subject can be. The most interesting example came from looking at segregation within cities and discussing how actually you can have people with a relatively high tolerance for having others, unlike themselves living around them, but still, over time, end up with a highly segregated population. Such unintuitive results are very clear to see once you write down a simple dynamical model of such a situation.

Anyway, another busy week beckons. This is going to continue unabated for the next three weeks, with a number of couchsurfers coming to visit along the way, until I head to Spain for a long weekend in March - I'm going to need the break by then...

No comments: