Zilong Zhong is a PhD student of Systems Design Engineering, University of Waterloo. He is supervised by Prof. Jonathan Li and Prof. Alexander Wong. His research interests include deep learning, computer vision, probabilistic graph models, and their applications to various data interpretation scenarios. On Feb. 14, 2018, he gave a seminar presentation with the tile "Supervised, unsupervised, and semi-supervised deep learning: what's next".
Supervised deep learning models, like convolutional neural networks (CNNs), have achieved unprecedented success in multiple image classification, localization, and segmentation tasks, thanks to the existing of three indispensable factors: large amounts of labeled data, reasonable computing power, and suitable deep neural networks. Unsupervised deep learning models, like generative adversarial networks (GANs), provide a solution to implicitly estimates real data distribution and correspondingly generates synthesized samples. Semi-supervised deep learning models integrate the discriminative deep learning models with their generative counterparts. In face of the explosion of artificial intelligence (AI), we should consider suitable deep learning models for different applications in the era of big data. In this seminar, three basic deep learning frameworks will be introduced in the context of hyperspectral image classification. Moreover, three promising AI research lines will be briefly discussed with examples, like image generation, image segmentation, and board games.
