Prof. Ali Shokoufandeh
Drexel University, Philadelphia, US
Approximation of hard combinatorial problems via embedding to hierarchically separated trees
Abstract: This talk deals with a bounds on several optimization problems including monochromatic and bichromatic optimization minimum matching, minimum spanning trees, and the travelling salesman problem. Many of these problems lend themselves to representations in terms of hierarchically separated trees --- trees with uniform branching factor and depth, and having edge weights exponential in the depth of the edge in the tree. We also present tight concentration results for the monochromatic problems. Finally, we apply our results to similar problems on random point sets taken from the Euclidean hypercube [0, 1]d.
Prof. David Hand
Imperial College, London, UK
Evaluating supervised classification methods: error rate, ROC curves, and beyond
Abstract: Evaluation lies at the core of supervised classification. Comparison between methods, selection of methods, parameter estimation, and feature choice all implicitly require the assessment of the effectiveness of classifiers. Because it is so important, a great deal of effort has been expended on developing performance measures and their estimators, stretching back over fifty years. But new applications, new computer power, and new inferential challenges mean that research continues. This talk explores some recent developments.