Eldan Goldenberg's lab notebook

GECCO braindump

I've spent most of the past week at GECCO - the Genetic and Evolutionary Computing Conference. It was, as expected, not as relevant to me as ALife, but at the same time it was gratifying to see how much quality content there was in the Real World Applications and Evolutionary Computation in Practice tracks. I should definitely make a point of going to the GECCO closest in time to my graduation, as it would be a very good way to find out about the sorts of jobs worth me applying to. Till then, I think I should prioritise the conferences with a more scientific and less engineering emphasis.

As a personal preference, I also prefer single-tracked conferences. The advantage of having multiple parallel tracks (GECCO had as many as 9 sessions going on at any given time) is that it allows greater breadth of content, but the drawback is that splinters the conference. I find that I get as much value out of a conference in terms of being able to talk to many like-minded people about their and my work as I do from the presentations themselves, and it's much easier to make conversation when everyone's been listening to the same thing, and is there because of a comparatively well-defined theme.

The customary braindump follows after the cut.

I'll do this one in two sections. First, a sampling of practical applications I heard about evolutionary computing techniques being applied to:

1. Diagnosing tumour metastasis from gene expression data
2. Clustering proteins by functional similarity
3. Deciphering the complex networks of interactions between proteins
4. VLSI layouts (this one was not news to me, as much of the inspiration for my MSc thesis came from VLSI layout work that had already been published by 2002)
5. Optimising the assignment of clothing cuts to a sheet of fabric, to minimise off-cut waste
6. At least two forms of automated financial market trading

And now a smattering of other thoughts:

7. My take-home message from the fitness landscapes tutorial was that many of the measures people typically use to try and get a handle on fitness landscape difficulty don't really capture anything important.
8. There were various interesting techniques presented for maintaining diversity in a population. I was most struck by the simplest one I heard: using the fairly-common 2d grid idea but adding a bias term that makes selection events much more likely in one dimension than the other. This has the advantage of needing little overhead to compute, and being easy to dynamically adjust through a run.
9. Simon McGregor had several good ideas for how to avoid overspecialisation in evolved agents; I'll come back to this in more depth in another post because it's a major issue for my own work
10. Related to the previous point: I think I finally got my head round exactly what pareto optimality is, and I have a little reading list to make sure
11. One of the keynote speakers presented the Allen Brain Atlas, which is a hugely impressive body of anatomical and gene-expression data about the mouse brain.