Neural Networks and Genetic Algorithms as Tools in Particle Physics
Rick Field - University of Florida
Tevatron University 6:30pm May 21, 1998
Genetic Algorithms (GA) are a broad class of minimization algorithms modeled after genetics and evolution. Unlike local algorithms, such as the gradient descent algorithm, GA's are much less likely to find and stay in a local minimum. This is a considerable advantage for a large class of problems, including many applications in Particle Physics. I will explain how GA's work and, as an example, use a GA to perform "optimal" multi-dimensional linear cuts. Six Modified Fox-Wolfram Moments, Hl, are used to characterize the event topology at the collider and a GA is employed to find the region in six-dimensional Hl-space that enhances signal over background for the six-jet decay of top-quark pairs. I will also explain how Neural Networks work and discuss how GA's can be used to improve the training of Neural Networks.

R. Field - May 4, 1998