Combinatorial optimization problems occur in a wide range of domains, from Bayesian network structure search to questions in neuroscience and biochemistry. However, all of these problems have in common the need to optimize some score, and often the calculation of this score is a significant source of slowness in the search for a solution. Through the use of a carefully calibrated approximation, however, this time can be significantly reduced with little effect on the quality of the results. I demonstrate here how such a proxy function can be used, as well as explore situations where the proxy strategy fails and offer reasons why or why not it might be suited to a particular problem.
bayesian, combinatoric optimization, optimization, machine learning, learning
Level of Degree
Department of Computer Science
First Committee Member (Chair)
Second Committee Member
Williams, Lance R.
Third Committee Member
Yackley, Benjamin. "Proxy-Based Acceleration for Combinatorial Optimization Problems." (2014). http://digitalrepository.unm.edu/cs_etds/42