Loading


Welcome to the Missouri Estimation of Distribution Algorithms Laboratory (MEDAL)
MEDAL focuses on the design, enhancement, analysis, and applications of estimation of distribution algorithms (EDAs), which represent a powerful class of stochastic optimization techniques inspired by evolutionary computation and machine learning. Using machine learning techniques in populationbased search allows automatic identification and exploitation of problem regularities, yielding robust and scalable performance on broad classes of challenging optimization problems.
An EDA starts with a population of random candidate solutions. In each iteration, a probabilistic model of promising solutions is first built; for example, the probabilistic model can be represented by a Bayesian network. The built probabilistic model is then sampled to generate new solutions, which are incorporated into the original population using some replacement strategy. The algorithm is terminated when given termination criteria are met; for example, the algorithm is stopped when a good enough solution has been found, when the population has lost diversity, or when a threshold on the number of iterations has been reached.
EDAs have been successfully used in a number of applications, including Ising spin glasses, MAXSAT, military antenna design, telecommunication network optimization, feature subset selection, supervised machine learning, and nurse scheduling.
Besides EDAs, MEDAL's research interests cover other branches of genetic and evolutionary computation, learning classifier systems, bioinformatics, and machine learning.
Missouri Estimation of Distribution Algorithms Laboratory 
Web design by Martin Pelikan 
