Deep Learning: The AI Revolution
Though neural nets aren't new which date back to the 1950s, we didn't notice the quantum leaps in the technology brought by deep learning until recent four or five years. The dramatic progress for deep learning is powered by the increased computational power and the accessibility to large and varied datasets. While researchers are attempting to develop the ability of deep neural networks to encode more and more human intelligence, the application of deep learning no longer highly centers around traditional supervised learning tasks such as image identification. Most recent new attempts to apply deep learning have broadened the boundary for deep learning applications across the tasks of interest for both academia and industry, such as creating agent that can draw or write like human with deep generative models, or autonomous driving and playing computer games with deep reinforcement learning. This talk will give an overview to deep learning by introducing the background, current breakthrough applications, and the prospect for the future.
Haiyan Yin is a PhD student and a research assistant at School of Computer Science and Engineering, Nanyang Technological University, working with Asst Prof Sinno Jialin Pan. She received her bachelor’s degree in Computer Engineering from Nanyang Technological University in 2014 with First Class Honors. Her current research centers around deep reinforcement learning, which integrates deep learning techniques with reinforcement learning to solve tasks in an end-to-end manner. Specifically, she is interested in transfer learning for deep reinforcement learning, where knowledge learned in certain task domains can be utilized to accelerate the learning of another new task via efficient domain adaptation.