Colloquia / Seminars / Industry Talks

In this talk, I will introduce our recent works in machine learning and computer vision. Among them, two works related to manifold alignment and deep learning will be presented in more detail: (1) We propose a generalized unsupervised manifold alignment method to build the connections between different but correlated datasets without any known correspondences. This method can be widely applied to real-world problems, such as alignment of face image sets across different appearance variations, structure matching of protein sequences, video face recognition and visual domain adaptation. (2) We propose deep network cascade, a cascade of multiple stacked collaborative local auto-encoders, for image super-resolution. Our method can gradually upscale low-resolution images layer by layer, each layer with a small scale factor, and achieve more promising results in visual quality as well as quantitative performance.

Hong CHANG received her PhD degree in Computer Science from the Hong Kong University of Science and Technology in 2006. She was a Research Scientist with Xerox Research Centre Europe. She is currently an Associate Researcher of the Institute of Computing Technology, Chinese Academy of Sciences. Her main research interests include algorithms and models in machine learning (e.g., metric learning, manifold learning, deep learning etc.), and their applications in pattern recognition, computer vision and data mining (e.g., object recognition, image super-resolution, video modeling, etc.).