Highlights from GECCO 2015

Posted July 28, 2015 by Emily Dolson in News / 0 Comments

I just got back from GECCO 2015. Despite my lack of ability to speak Spanish (it was in Madrid), I had a great time, and I was really impressed by a number of the talks. So here’s a quick highlight reel:

  • All three keynotes were great, but the first one, by Ricard Sole, was my personal favorite. It was about synthetic biology, which is normally something I only care about the very high level details of, but the applications in this talk were really exciting! He discussed engineering new ecosystems to replace those we are damaging beyond repair, in case of emergency. The two primary examples he discussed were the human gut microbiome and macro-scale ecosystems that we are losing to climate change. Since this was GECCO, he didn’t talk too much about the ecology side, but I asked if he thought there was still ecological theory that we need to understand in order to make this work, and he said there is. Since I’ve often thought of these ideas as one of the long-term applications of my research, that was really exciting to hear!
  • Jeff Clune gave a fantastic talk on his new Innovation Engine algorithm, in which a deep neural network is combined with novelty search to create images that vary at an abstract and interesting level. The images generated by this algorithm were so convincing that they received an award at an art show without being identified as computer-generated. I particularly liked the concluding statement that “perhaps the problem with objective functions is just having too few of them.” Here are some example images from the paper, which received a best paper award for the ALIFE+GDS track:

    Images generated by the Innovation Engine Algorithm (Nguyen, Yosinski, and Clune, 2015)
    Images generated by the Innovation Engine Algorithm (Nguyen, Yosinski, and Clune, 2015)

 

  • In another talk (that won best paper), “Mk Landscapes, NK Landscapes, MAX-kSAT: A Proof that the Only Challenging Problems are Deceptive,” Daryl Whitley brought together theory on NK landscapes and MAX-SAT problems. I particularly liked the point that, despite the frequency with which we use random benchmarks, interesting real-world problems are basically never random.
  • In the same session, Brian Goldman gave a great talk (which was also nominated for best paper) in which he presented a version of his Parameterless Population Pyramid (P3) algorithm designed to work on grey box problems.
  • Justin Pugh, a grad student in Ken Stanley‘s lab, presented a paper proposing a benchmark problem for quality diversity algorithms. I’m really excited by the potential future applications of this work, mostly because I’ve been kind of frustrated with the lack of comparison of diversity algorithms. This paper also dealt with the question of what properties of a novelty metric make novelty search effective. As an added bonus, they’ve got cool visualizations.
  • Joel Lehman gave an interesting talk on using extinction in evolutionary computation. I admit I was a little unconvinced by an earlier talk I saw him give on this subject, but this one definitely made up for that. The main point was that periodic mass extinctions can select for evolvability, but only when paired with a “divergent” search method, i.e. one that diversifies rather than converging on a specific solution. He may or may not have also been making the point that extinctions can result in more specific niches and thus greater diversity (something which I vaguely recall there being evidence for in the fossil record, although I can’t find a citation at the moment so take that claim with a grain of salt).
  • It was great to see this comparison of novelty search algorithms, given my aforementioned frustration with people not comparing similar algorithms to each other.
  • Women@GECCO put together a fun slate of activities, including a “science slam” in which participants presented their research in various creative ways. The two most notable of these involved a presentation made completely in rhyme and a demonstration using duplos. This session also featured a great keynote by Emma Hart, in which she compared the trajectory of her research to a mountainous road in Scotland. Following all this, there was the work/life balance discussion that Anya, Nur Zincir-Heywood, and I co-organized. I like to think it went well :). More on that in a future post!

 

There was a lot going on at GECCO, so I’m sure I missed a bunch of awesome things. What were your favorites? Let us know in the comments!

Emily Dolson

I’m a doctoral student in the Ofria Lab at Michigan State University, the BEACON Center for Evolution in Action, and the departments of Computer Science and Ecology, Evolutionary Biology, & Behavior. My interests include studying eco-evolutionary dynamics via digital evolution and using evolutionary computation techniques to interpret time series data. I also have a cross-cutting interest in diversity in both biological and computational systems. In my spare time, I enjoy playing board games and the tin whistle.

More Posts - Website - Twitter

Leave a Reply