主讲人: Galen Papkov 博士，UIC

时间: 2011年9月21日（星期三） 下午3:30-4:30

地点: E304

摘要:

Smoothed polynomial histograms attain their computational efficiency by generating a nonparametric density estimate that attempts to match bin moments. This work improves upon the smoothed polynomial histogram by incorporating inequality constraints corresponding to confidence intervals for the local sample moments. The use of confidence intervals provides increased adaptivity. Smoothed polynomial histograms can be applied to pre-binned and massive data sets. In addition to density estimation, applications include bump hunting, classification, and change-point analysis. Future work will explore the effects of adaptive knot selection and higher-order derivatives in the penalty on the quality of the density estimate.