reganbaucke/sprocket.jl

A framework for stochastic programming in Julia

Sprocket.jl

Sprocket: (S)tochastic (PRO)gramming

Sprocket is a framework for describing stochastic programmes in an algorithm-agnostic fashion.

Once a problem has been described, the user can implement an algorithm to solve the problem, or she can use an algorithm that we've prepared earlier. Sprocket was developed to compare the speeds of different solution algorithms in a fair and consistent manner. Besides this, Sprocket is a great way to write and solve your stochastic programme!

Problem Class / Decomposition

Sprocket solves stochastic optimisation problems where uncertainty appears in the objective function. Currently, there is no support for optimisation problems with chance constraints or with random feasible sets.

Bugs

Please raise an `issue' if you experience an error while using Sprocket.

sprocket.jl's Issues

Providing a default implementation for SAA Algorithm

It would be useful to include the SAA algorithm for solving stochastic optimisation problems as a default algorithm for Sprocket.

Obtaining an upper bound for problems with unbounded random variables

Currently, problems with bounded support for the random variables are the only problems which can be solved. By using the theorem from the paper, we can still obtain an upper bounded for these problems, provided the functions growth is bounded somehow (i.e. Lipschitz-continuous functions). We would like to implement this feature.

Add control problem

Currently, we can only solves problems involving the expectation of a convex function.
We need to add the ability to solve stochastic programmes.

Updating bound information in the algorithm

In the current implementation, we can only control the running the algorithm by the number of iterations. We would like the default algorithm to update the abs_tol and rel_tol fields of the criteria field that the algorithm carries in its state.