It's the first day of Unconventional Computation and Natural Computation 2023 conference.
It started off with a great keynote by Thomas Bäck, titled "On the automatic optimisation of problem-specific optimisation heuristics gleaned from nature". Many population-based optimisation algorithms are essentially variations on a theme, with different parameter settings. Rather than invent yet another variant, instead, optimise the parameter settings for your particular application. This can lead to "weird" algorithm configurations, but better ones than we can design by hand. Then, what to do if your cost function is too expensive to calculate for this approach? Then use a suitable cheaper proxy for the tuning, and use the tuned algorithm on the real problem. However, standard benchmarks aren't good proxies, so use a random function generator, and choose results that are good proxies. (And there was more good stuff too, but noting it here would make this post too long: go read his papers instead!)
Next was supposed to be the first half of the TEMC (Theoretical and Experimental Material Computing) workshop that I was running. Unfortunately, my invited speaker was not having the same smooth travel experience from Canada that I had experienced from the UK, and wasn't going to arrive until lunchtime. So I liaised with the conference organiser to swap some talks around.
So instead, we had a session of conference presentations by my team, on quantum computing and reservoir computing.
After lunch, Ion Petrie delivered an invited tutorial on Reaction Systems. I had come across these before, but it was good to have a session putting everything in one place.
Last but not least, my workshop speaker, Julien Sylvestre, gave a fascinating invited talk on unconventional robotics: using unconventional computational models in concert with unconventional hardware.
So, a great opening to the conference. Off to the evening welcome reception at a beach-front venue.
No comments:
Post a Comment