【数学学院】Data Adaptive Support Vector Machine with Application to Prostate Cancer Imaging Data

报告题目：Data Adaptive Support Vector Machine with Application to Prostate Cancer  

Imaging Data

报告人：Wenqing He

报告人单位：University of Western Ontario

报告时间：7月10日（周二）下午14:30-15:30

报告地点：数学学院西303报告厅

邀请人：周杰

Abstract:

Support vector machines (SVM) have been widely used as classifiers in various settings including pattern recognition, texture mining and image retrieval. However, such    

methods are faced with newly emerging challenges such as imbalanced observations and 

noise data. In this talk, I will discuss the impact of noise data and imbalanced 

observations on SVM classification and present a new data adaptive SVM classification 

method.This work is motivated by a prostate cancer imaging study conducted in London

Health Science Center. A primary objective of this study is to improve prostate 

cancer diagnosis and thereby to guide the treatment based on statistical predictive 

models.The prostate imaging data, however, are quite imbalanced in that the majority 

voxels are cancer-free while only a very small portion of voxels are cancerous. This 

issue makes the available SVM classifiers typically skew to one class and thus 

generate invalid results. Our proposed SVM method uses a data adaptive kernel to

reflect the feature of imbalanced observations; the proposed method takes into 

consideration of the location of support vectors in the feature space and thereby 

generates more accurate classification results. The performance of the proposed method is compared with existing methods using numerical studies.