We want to start a new bachelor project soon, with the goal of using Genetic Programming to somehow predict equations of motion from timeseries. There has been quite some work on this field, notably the publications
- Distilling Free-Form Natural Laws from Experimental Data, M. Schmidt and H. Lipson, Science 2009
- Automated reverse engineering of nonlinear dynamical systems, J. Bongard, H. Lipson, PNAS 2007
- The book by Koza : “Genetic Programming: On the Programming of Computers by Means of Natural Selection”
I was wondering whether the community has more suggestions, or if the author (Richie Lee) has some comments on whether this package is a good place to start. My only fear is that ExprOptimization.jl is under-documented, which will make things hard for a bachelor student, or anyone who would like to expand the package, really.
In addition, the speed seems to be on the slow side as well, as I have been running the example given in the notebook for more than 10 minutes now and it still hasn’t finished computing. Of course I have to say that I have never used Genetic Programming before, so I do not have any clue on how fast these things should be!
In general we would like to create something that can be incorporated into the DynamicalSystems.jl ecosystem and the two goals we have are:
- Predict system conserved quantities using Julia
- Predict or at least indicate form that equations of motion might have, again using Julia.
P.S.: If this topic is better suited in another category, please do not hesitate to move it.