If nonlinear mathematical optimization counts as machine learning, there's been at least one paper (see my other post[1] and optimization group's page[2]). Car

If nonlinear mathematical optimization counts as machine learning, there's been at least one paper (see my other post[1] and optimization group's page[2]). Carpalx[3] is more like what you envision, but it uses simulated annealing instead of mathematical programming[4], so you won't know how far from the true optimum its solutions are. [1] https://news.ycombinator.com/item?id=27428539 https://news.ycombinator.com/item?id=27428539 [2] http://resources.mpi-inf.mpg.de/keyboardoptimization/ http://resources.mpi-inf.mpg.de/keyboardoptimization/ [3] http://mkweb.bcgsc.ca/carpalx/ http://mkweb.bcgsc.ca/carpalx/ [4] http://mkweb.bcgsc.ca/carpalx/?simulated_annealing http://mkweb.bcgsc.ca/carpalx/?simulated_annealing