July 9-13 multiple conferences

[dmm]

38th International Conference on Mathematical Foundations of Programming Semantics: www.cs.cornell.edu/mfps-2022/ (July 11-13; virtual participation was 10 dollars last time I checked)

The Genetic and Evolutionary Computation Conference, July 9-13 (Boston and virtual, I'll be attending virtually): gecco-2022.sigevo.org/HomePage (the least expensive way to attend virtually is to become an ACM SIGEVO Member, it's only $25, and then the attendance is free).

Sparsity in Neural Networks

Advancing Understanding and Practice

SNN Workshop 2022 (July 13, virtual, free registration): www.sparseneural.net/
 

Comments

[dmm]

The previous sparsity workshop: https://dmm.dreamwidth.org/45791.html

[dmm]

GECCO is huge, multitrack, and quite messy in terms of technical glitches and organization so far. Nevertheless, plenty of good stuff is hidden there (e.g. Risto Miikkulainen's "Evolution of Neural Networks" tutorial is quite amazing; I have no idea if there would be recordings of all this). The program booklet is 160 pages, I have a copy of that.

[dmm]

Here it is: https://www.cs.utexas.edu/users/risto/talks/enn-tutorial/

[dmm]

And this is his answer with respect to AI safety for evolutionary algorithms: "If you are evolving solutions for a real-world task where there is a possibility of harm (e.g. investments, treatments), you would do it with a surrogate first, and then make sure the solutions are safe before deploying them---and it is important that such solutions are explainable"

Actually, it is becoming popular to consider interplay of evolutionary methods and explainability (this is rather counter-intuitive as normally the evolutionary solutions end up being rather opaque, but one can explicitly put evolutionary pressure to make things more explainable)

[dmm]

Other tutorials of the first day I have been attending:

Generative Hyper-heuristics (Tauritz,Woodward)

Automated Algorithm Configuration and Design (López-Ibáñez, Stützle, Pérez Cáceres) (includes "iterated racing" + AutoMOEA auto multi-objective evolutionary algorithms)