Eric Mjolsness, "Toward a Searchable Space of Dynamical System Models"

Posted by Kathryn Cramer.

Mjolsness set out criteria for simple formal systems: expressive power, discrete labeled graph structure, & self-applicability. he showed a wide variety of examples from a number of disciplines and on multiple scales. He argued that the stochastic process operators can form the semantics of a language, and that operator algebra then leads to novel algorithms. This results in a searchable space of simple dynamical system models.

Abstract. Also, here is Eric Mjolsness's web site, which he  suggests you visit for more information, is here.

Comments

[Kovas Boguta]

This talk was interesting, though the pace was difficult to keep up with.

It seems he was come up with a definite computational system, which is capable of a lot of things. Sort of a computational primitive that spans a wide class of models common in chemical-like systems and elsewhere.

I think there is room for making it simpler. I couldn't really tell which of the components were essential, though Eric said there were basically three parts to the framework, so it doesn't sound complicated either.

Some of the features of the framework were clearly motivated to address a spectrum of existing problems. I wonder how it can be simplified, and explored for its inherent computational behavior.

In any case, interesting.