About the program:. The identification of useful biomarkers is becoming a common problem for assessing a drug response, treatment outcome, or adverse side effects. Typically, the biological data sets are small and have potential features that are far in excess of the number of observations. This problem is often in the form of a case-control study and analyzed as a set of univariate markers, employing multiple hypothesis testing controls. However, this problem can also be cast as a feature subset selection and statistical pattern recognition problem. While both filter and wrapper based feature subset selection methods may be appropriate for this problem, care must be taken in their application, due to various over fitting issues that can lead to biased performance estimation and poor generalization. A discussion of Kernel based feature selection tailored to small data set problems is presented with an application to the pharmacogenetics of calcium channel blockers treatment outcome in hypertension.
About
the Speaker: Troy
Bremer received his B.S. degree from
Time/Place: Wednesday Jun
14, Wednesday 6:00 P.M. Lockheed Martin, 4770 Eastgate
Mall
Free for IEEE members, $5 otherwise.
Reservations/Information: Andrew Diamond (IEEE CIS San Diego Chapter Chair) (858) 509-3115, adiamond@EnvisionSystemsLLC.com