Topic：Recent Advances in Deep Learning for Image Classification

Date&Time：2015/11/27(五) 14:20

Speaker： 詹宗翰 博士 (聯發科)

Location：清華大學台達館 R216

Abstract :
Machine learning is everywhere nowadays, and large-scale machine learning amounts to numerical optimization of weights for human designed representations and features. The goal of deep learning is to explore how computers can take advantage of huge amount of data to develop features and representations appropriate for complex interpretation tasks. The talk aims to cover the basic motivation, ideas, models and learning algorithms in deep learning for image classification. These methods have been shown to perform well on various image classification tasks, including Large Scale Visual Recognition Challenge on ImageNet. The most attractive advantage of these techniques is that they can perform very well without external hand-designed resources or time-intensive feature engineering. The first part of the talk presents brief history of neural networks and an overview of deep learning in industry. The second part involves basics of neural networks and training algorithms via back-propagation. In the last part, a subfield of deep learning, convolutional neural networks and its application in image classification will be introduced. The focus of this talk is to provide insight and understanding, with the hope that the messages to be delivered could bring out some more interesting ideas in audiences' researches.

Bio :
Tsung-Han Chan received the Ph.D. degree from the Institute of Communications Engineering, National Tsing Hua University (NTHU), Hsinchu, Taiwan, in 2009. He is currently working as a Senior Engineer with MediaTek Inc., Hsinchu, Taiwan. Before joining MediaTek, he was a visiting Doctoral Graduate Research Assistant with Virginia Polytechnic Institute and State University, Arlington, in 2008, and a Postdoctoral Research Fellow with NTHU from 2009 to 2012, and a Research Scientist at Advanced Digital Sciences Center, Singapore from 2012 to 2013. He was also a Staff Engineer at Sunplus Technology Corp. in 2014.

Dr. Chan was awarded National Science Council Graduate Student Study Abroad Program (補助博士生赴國外研究) and CTCI Foundation Science and Technology Research Scholarship (中技社科技研究獎學金) during his PhD program in NTHU. His PhD dissertation also received the Best Doctoral Dissertation Award of CID & ORSTW (中華決策科學學會暨台灣作業研究學會) and Outstanding Doctoral Dissertation Award of IPPR (中華民國影像處理與圖形識別學會). He was also awarded National Science Council Postdoctoral Research Abroad Program (補助赴國外博士後研究) in 2012.

Dr. Chan was a co-recipient of a WHISPERS 2011 Best Paper Award. He was recognized as an Outstanding Reviewer for IEEE Computer Vision and Pattern Recognition (CVPR) 2014, and also recognized a Best Reviewer of the IEEE Transactions on Geoscience and Remote Sensing (TGRS) 2014. He has published more than 50 international technical papers with 785 Google citations and 12 h-index.

His research interests are in image processing, numerical optimization and machine learning, with a recent emphasis on deep learning, computer vision, and hyperspectral remote sensing.