【電機系書報討論演講訊息】

書報討論網頁：http://www.ee.nthu.edu.tw/~ee591000/

演講公告網址：/p/403-1175-2629-1.php

題   目：Unsupervised Learning via Topic Models

講   員：簡仁宗 教授 (國立交通大學)

時     間：2014年6月13日(星期五) 下午2時20分 (2時10分入場)

地         點：台達館105教室

Abstract

In this talk, I will introduce a series of latent topic models which are developed for unsupervised speech and language processing. In general, latent topics are extracted to reflect the unknown clusters and explore the underlying structure from a set of observation data through an unsupervised procedure. However, the observation data are usually heterogeneous and significantly scaling up in the era of big data. It becomes crucial to build a hierarchical, adaptive, scalable and flexible topic model. Considering this perspective, I will survey a series of solutions, which are implemented by using approximate Bayesian inference based on variational Bayesian and Gibbs sampling, and point out possible extensions for speech and language applications.

~歡  迎  聽  講~