好文作者面授招–20151021

【15-33期VALSE Webinar活动】

报告嘉宾:谢澎涛(CMU ML Dept)
报告时间:2015年10月21日(星期三)晚20:00(北京时间)
报告题目:Diversity Regularization of Latent Variable  Models: Theory, Algorithm and Applications [Slides]
主持人:禹之鼎(CMU ECE Dept)
报告摘要:Latent Variable Models (LVMs) or latent space models are a large family of machine learning models that have been widely utilized in computer vision, text mining, computational biology, to name a few. In this talk, I will introduce a new type of regularization approach of LVMs: diversity regularization, which encourages the components in LVMs to be diverse, with the aim to (1) capture long tail factors in knowledge; (2) reduce model complexity without sacrificing expressiveness. Specifically, I will introduce the motivation of designing this regularizer, how it is formally defined, how to optimize it, its theoretical analysis and empirically applications.
参考文献:
[1] Pengtao Xie, Yuntian Deng, and Eric Xing. Diversifying Restricted Boltzmann Machine for Document Modeling. In ACM SIGKDD Conference on Knowledge Discovery and Data Mining. (KDD 2015)
[2] Pengtao Xie. Diversifying Distance Metric Learning. In European Conference on Machine Learning. (ECML 2015)
报告人简介:Pengtao Xie is a PhD student in the Machine Learning Department at Carnegie Mellon University. His primary research focus is latent variable models, in particular developing geometric regularization approaches to reduce model complexity without compromising expressiveness and building distributed systems to facilitate large scale latent variable modeling. He received a M.E. from Tsinghua University in 2013 and a B.E. from Sichuan University in 2010. He is the recipient of Siebel Scholarship, Goldman Sachs Global Leader Scholarship and National Scholarship of China.

报告嘉宾:卢策吾(Stanford University)
报告时间:2015年10月21日(星期三)晚21:00(北京时间)
报告题目:Two-Class Weather Classification [Slides]
主持人:章国锋(浙江大学)
报告摘要:Image weather understanding is a relatively new topic in computer vision. Different from the important object recognition and classification problems, weather  recognition needs to understand more complex phenomena. In this talk, I will present the overview and motivation of weather understanding and my work on it. I will mainly discuss the models for two-class weather classification and smog visibility estimation. I also will introduce two large-scale weather datasets we have built. Finally, future work will be discussed.
参考文献:Cewu Lu, Di Lin, Jiaya Jia, Chi-Keung Tang, “Two-Class Weather Classification”, IEEE Conference on Computer Vision and Pattern Recognition, (CVPR), 2014.
报告人简介: 卢策吾,2013年获香港中文大学博士学位,现为斯坦福大学人工智能实验室(Stanford University, AI lab)博士后研究员(research fellow)。其研究兴趣包括物体识别,动作检测,图像与文本联合分析。目前发表(含接收)18篇CCF-A类文章,其中包括11篇CVPR/ICCV, 一篇提出算法在open CV中应用,一篇为SIGGRAPH ASIA 近5年内被引用数最高。作为领队,参加 ImageNet比赛 ILSVRC 2014 获 object detection track (without extra data)第四名。

(Visited 628 times, 1 visits today)