 | From the Series از مجموعه : Random Matrices and Spectral Clustering Abstract
Produced by تهيه كننده : Microsoft Research
Date تاريخ : 2006-07-27 Random Matrices and Spectral Clustering Abstractdownload دانلود ,ويدئو و اسلايد Video & Slide , از گروه Computer Sience & Engineering كامپيوتر و مهندسی كتابخانه اينترنتي دانش گستران جوان You Research Description توضيح : Theoretical analysis of Spectral Clustering is often based on the theory of random matrices which assumes that all entries of the data matrix (with each row representing a data object) are independent random variables. In practice, however, while the objects may be independent of each other, features of an object are not independent. To address this, we will prove bounds on the singular values of a matrix assuming only that its rows are independent vector-valued random variables (the columns may not be independent) and describe a new clustering algorithm with this limited independence.
In the second part, we address a more theoretical question. We will show a generalization of Azuma's (martingale) inequality to the case when the random variables are matrices and more generally infinite-dimensional operators.
The first part is joint work with A. Dasgupta, J. Hopcroft and P. Mitra.
Related Links لينكهای مرتبط : - Random Matrices and Spectral Clustering Abstractdownload دانلود ,ويدئو و اسلايد Video & Slide , از گروه Computer Sience & Engineering كامپيوتر و مهندسی كتابخانه اينترنتي دانش گستران جوان You Research Speaker(s) اجرا : Ravi Kannan, Ph.D., Theoretical Computer Science, Applied Mathematics, Optimization and Discrete Mathematics, Yale University
Runtime مدت زمان : 00:55:46
Video Size حجم ويدئو : 122 MB
Number of Slides تعداد اسلايدها : 78 (8 MB) Random Matrices and Spectral Clustering Abstractdownload دانلود ,ويدئو و اسلايد Video & Slide , از گروه Computer Sience & Engineering كامپيوتر و مهندسی كتابخانه اينترنتي دانش گستران جوان You Research
|