报告题目：Is There a General Theory for the Detection of Anomalies in Images?
报告人： Jean-Michel Morel 教授 法国加香高师
主持人： 沈超敏 副教授
Anomaly detectors address the difficult problem of detecting automatically exceptions in an arbitrary background image. Detection methods have been proposed by the thousands because each problem requires a different background model. By analysing key examples of the literature, we show that all anomaly detectors are characterized by their choice among seven fundamental principles guiding the background model and the decision method. We show that these principles can be combined in a general method that uses six of them. Our synthesis reduces the problem to the easier problem of detecting anomalies in noise. In that way, the varifold background modeling problem is replaced by simpler noise modeling, and allows the calculation of rigorous thresholds based on the a contrario detection theory. Our conclusion is that it is possible to perform automatic anomaly detection even on a single image. (Joint work of Thibaud Ehret, Axel Davy, Jean-Michel Morel, Mauricio Delbracio)
Jean-Michel Morel, professor at the École normale supérieure of Cachan. Mathematics Center and their applications, lauréat of the Grand Prix Inria - Academy of Sciences in 2013. Focusing on the analysis and mathematical processing of images, Prof Morel’s most notable contributions are in the areas of segmentation, denoising, mapping, and detecting significant events in digital images. Prof. Morel is the founder of the online scientific publication “Image Processing OnLine” (http://www.ipol.im/). He has won numerous prizes, including Philip Morris Mathematics Prize (1991), CISI-Engineering Award for Applied Mathematics (1992),Science and Defense Award (1996), INRIA Grand Prix - Academy of Sciences (2013), CNRS Medal of Innovation (2015) and IEEE CVPR Longuet-Higgins Prize (2015).
报告题目：Deep Approximation via Deep Learing
报告人： 沈佐伟 教授 新加坡国家科学院院士, 新加坡国立大学
主持人： 沈超敏 副教授
The primary task of many applications is approximating/estimating a function through samples drawn from a probability distribution on the input space. The deep approximation is to approximate a function by compositions of many layers of simple functions, that can be viewed as a series of nested feature extractors. The key idea of deep learning network is to convert layers of compositions to layers of tunable parameters that can be adjusted through a learning process, so that it achieves a good approximation with respect to the input data.
In this talk, we shall discuss mathematical foundation behind this new approach of approximation; how it differs from the classic approximation theory, and how this new theory can be applied to understand and design deep learning network.
沈佐伟教授，新加坡国立大学理学院院长，陈振传百年纪念教授，新加坡国家科学院院院士，美国数学会会士（AMS Fellow), 美国工业与应用数学会会士（SIAM Fellow)。主要研究领域为逼近与小波理论、时频分析、图像科学， 学习理论等。作为国际著名数学家，沈佐伟教授先后获得Wavelet Pioneer奖、新加坡国立大学杰出科学研究奖和新加坡科学成就奖，并受邀在2010年国际数学家大会和2015年国际工业与应用数学大会上作报告。