This is a significant generalization of my earlier stochastic optimization efforts. I presented as an author at the INFORMS 2015 Annual Conference in Philadelphia.

Slides

A big part of the generalization was defining a more abstract setting for the problem. In a nutshell, this work concentrates on two things:

  1. Using interval estimation to provide confidence levels and error tolerances.
  2. Attempting to optimize over continuous and discrete domains.

Some of the notes created during the research process for both deterministic and stochastic cases: PDF1, PDF2

The experimental results from the presentation were generated and processed in Python before using Mathematica to render the graphs.

Figure 1: Samples required as a function of both confidence level (\\(\alpha\\)) and tolerance.

Code

This is an ongoing work, so expect to see another post about this before 2016!