In his talk,
Hans Paul Schwefel has commented the past, the present and the future of evolutionary optimization. The growth of the field has been almost exponential, but regarding the future, what can only be said is that the unexpected should be expected.
He's spent most of his keynote talking about the past, and specially, what made him veer away from traditional optimization (he's mentioned a paper reviewer comment on "why should other optimization algorithms be necessary") to evolutionary optimization, specially in the first half of the seventies, when he started to apply evolutionary techniques (what would be called Evolution Strategies) to nozzle optimization.
There were several challenges: come up with a self-adaptive variance, which he solved by getting inspiraton from nature, parallelism, which has been solved in several ways: predator-prey or parallel cellular GA, and, finally the handling of constraints, which is hard to do in ES.
He's commented that evolutionary algorithms are getting less bio-inspired in time, this is not good or bad, but it's more interesting for him to look at models than to have super-tweaked ultra-tuned purportedly bio-inspired algorithms.
All in all, and as always, an interesting and insightful talk. A good headstart for the conference.
