2

i am looking for solver for solver/optimizer for markov decision process / stochastic optimal control problem (see also Sequential Decision Making under Uncertainty.

The problem is described by set of differential equations but it my by discretized in time.

d w / dt = u f(w) + z * w
d R / dt = (1-u) g(w)

where f(.),g(t) is some function, z is a random variable with normal/log normal/power law probability distribution, u is a control which is contained in [0,1]. u may change in time.

Optimization criteria may be mean revenue in the final time T:

max_u E[R(T)]_z

I am looking for library, something similar to http://nicky.vanforeest.com/probability/mdp/mdp.html but in c/c++, that solve this kind of problems.

jasiek
  • 21
  • 3
  • I don't know who up-voted this question, even though he didn't show any sign of effort.He just ask us to write the code in `"c/c++"` even if `c` and `c++` are different languages. – bitcell Nov 03 '14 at 13:47
  • "solver/optimizer" i.e. library which implements routines that solve this type of problems (like solver for PDE, eigenvalue problems etc...) not some one who solve the problem. – jasiek Nov 03 '14 at 17:30

0 Answers0