报 告 人：陈付华 教授，美国西自由大学
邀 请 人：徐凌宇 教授
A fundamental technology of artificial intelligence is pattern recognition. A basis of pattern recognition is classification. The key of classification lies in metric design. Due to its intrinsic drawbacks of Euclidean distance in classification, the Mahalanobis distance has been widely used in recent decades. Traditional classification methods always compute distances between instances. Different from traditional methods, this research takes each class as a distribution and computes between-class distances using information geometry. Under some assumptions, the average within-class distance among the same class is proportional to the standard deviation (for a random variable) or the product of standard deviations of each feature (for a random vector), and the between-class distance coincides with the Mahalanobis distance. The method is then applied to person re-identification, which is a very important application in a 5G time, such as smart city. To our surprise, the proposed method is very competitive compared with many state-of-the-art methods while saving the computational cost in the learning stage. Experimental results demonstrate the effectiveness of the proposed method.
陈付华， 南京大学理学学士，南京理工大学工学博士，佛罗里达大学(University of Florida)应用数学博士。现为美国西自由大学（West Liberty University）自然科学与数学系终身教授。