研究生招生

统计学院2018年博士研究生招生简章

编辑:时间:2017-08-24 19:31:03 浏览次数:

On April 28th, Professor Kong Linglong from the Department of Mathematics and Statistical Sciences, the University of Alberta in Canada, was invited to give a live online lecture entitled Significant Anatomy Detection Through Sparse Classification: A Comparative Study (a comparative study of important structural inspection based on sparse classification) for teachers and students of the School of Statistics of our school. The lecture was presided over by Professor Liu Xiaohui, deputy dean of the School of Statistics, and was watched by more than 60 teachers and students from the School of Statistics.


At the meeting, Professor Kong made a wonderful exposition on the application of statistics in neuroimaging research. First of all, Professor Kong shared the successful experience of cross-industry cooperation in the field of statistics and neuroimaging data analysis with students based on his own experience. Professor Kong mentioned that in order to fully understand the development direction, cutting-edge trends and research issues of a field, participating in relevant academic conferences and submitting papers to top academic conferences in the industry is the best way. By understanding the research topics of other professionals in the field, and using the comments and feedback of authoritative persons in the field, researchers can usually have a comparatively deep-going understanding of the new field. Secondly, Professor Kong pointed out that an important concern in the field of neuroimaging is to distinguish the patient from the normal population by analyzing brain scan image data, which has broad application prospects in the diagnosis of Alzheimer's patients in the medical field. One of the existing mainstream analysis methods is to analyze the image data as a combination of multiple independent small cubes, but this method ignores the functional correlation between the adjacent areas of the brain, which will lead to certain errors in the analysis results. The second method is to recognize different patterns of image data by using high-dimensional data classification models. However, this method requires a large sample size, and it is often unable to meet the needs of this sample size in real life. Based on this, Professor Kong’s team proposed a new type of sparse classification method. The method constructs a corresponding objective function by adding an image-based penalty item and theoretically gives the bounds of the estimated parameters; at the same time, considering that the existing objective function is a non-smooth convex function, which is difficult to calculate in a certain degree. Therefore, the article considers using the ADMM algorithm to split the existing objective function into a smooth and derivable optimization function and a non-smooth but show expression. In the part, the L-BFGS algorithm and the soft thresholding operator method are used to solve the two parts separately. Finally, Professor Kong showed the specific performance of the four types of candidate models on simulated data and empirical data and gave the best classification model.


In the course of the lecture, Professor Kong cited classics and interspersed anecdotes from relevant statisticians when referring to famous algorithms, which made the atmosphere of the lecture lively and interesting. Professor Kong’s report combines theory with practice, greatly improves the legibility of statistical theoretical knowledge, broadens the horizons for college teachers and students to engage in research in related fields, stimulates students’ interest in learning, and enhances students’ interest in learning Understanding of the frontier research direction of statistics.

 

Text: Zhu Ge

Picture: Zhu Ge

 

 

[Extended Reading] The University of Alberta is one of the largest research universities in Canada, and its research atmosphere and research conditions enjoy a widespread reputation in Canada and even North America. The University of Alberta alumni include the 16th Prime Minister of Canada, three Nobel Prize winners, 75 Rhodes Scholars, and 111 Canadian Chief Research Professors. Its artificial intelligence major is in a leading position in the world. Rich Sutton, the father of reinforcement learning, and Alpha Go's main authors David Silver and Aja Huang are all from the University of Alberta. As a young and middle-aged representative of Canadian statistics, Professor Kong's main research fields are robust statistics, functional data, statistical machine learning, and neuroimaging data analysis. Nearly 40 accepted papers have been published in the top international journals in the field of statistics, The Annals of Statistics, Journal of American Statistical Association, Journal of the Royal Statistical Society: Series B and other journals. Professor Kong is currently an editorial board member of Journal of American Statistical Association and The Canadian Journal of Statistics, and has extensive cooperation with other related scholars. Professor Kong currently has a large academic team consisting of visiting scholars, postdoctoral fellows and PhD candidates, including 2 members in the School of Statistics of Jiangxi University of Finance and Economics.