US-France (INRIA) Cooperative Research: Robust Semi-Supervised Clustering with Application to Multi-Modal Database Categorization

[405166frigui]孟菲斯大学的Hichem Frigui研究小组和法国国家信息学与应用数学研究所（INRIA） Nozha Boujemaa领导的IMEDIA研究小组之间的这项美法合作研究项目，重点是开发适用于对大量多模态科学数据收集进行分类的有效聚类算法。该提案从理论方面阐述了聚类算法及其在分析和组织科学数据集中的应用，即：(1)近50年来植物生物多样性数据的科学文本和相关图像；(2)拟南芥基因组计划基因表达数据。智力优势：该项目将推进使用模糊集理论和统计估计器集成的数据聚类方法的知识。他们将开发新的算法来识别子空间中的数据簇，结合多模态特征，并使用少量标记样本来指导聚类过程。更广泛的影响：新的算法以及科学数据集将与美国研究人员在其NSF-CAREER项目资助下的基于内容的图像检索（CBIR）原型相结合。CBIR原型将用于对工科和计算机专业大一学生以及高中学生的教育活动。所开发的方法将对信息安全、生物信息学、基于内容的多媒体和其他大型数据集的应用有用。通过该奖项，美国学生有机会在国际研究环境中发展研究技能，并为未来的合作与法国研究人员建立伙伴关系。

0405166FriguiThis U.S.-France cooperative research project between Hichem Frigui's research group at the University of Memphis and the IMEDIA research team led by Nozha Boujemaa at the French National Institute for Research in Informatics and Applied Mathematics (INRIA) in Rocquencourt focuses on the development of effective clustering algorithms suitable for categorizing massive multi-modal scientific data collections. The proposal addresses theoretical aspects of clustering algorithms and their applications in analyzing and organizing scientific data sets, namely: (1) scientific text and related images in botanical biodiversity date produced in the last 50 years; and (2) gene expression data from Arabidopsis thaliana genome project.Intellectual Merit: The project will advance knowledge in data clustering methods using integration of fuzzy set theory and statistical estimators. They will develop new algorithms that identify clusters of data in subspaces, that combine multi-modal features, and that use few labeled samples to guide the clustering process.Broader Impacts: The new algorithms, along with the scientific data sets, will be combined with the U.S. investigator's content-based image retrieval (CBIR) prototype funded under his NSF-CAREER project. The CBIR prototype will be used in education activities for freshman engineering and computer science students as well as high school students. The methods developed will be useful for applications in information security, bioinformatics, content-based multimedia and other large data sets. Through this award, U.S. students are given the opportunity to develop research skills in an international research environment and initiate partnerships with French researchers for future collaboration.