Deep Neural Networks for Lifelong Learning and Goal-Driven Computer Vision with Chris Kanan

November 20

12:00pm - 1:00pm

Engineering & Computer Science, Room 410, Classroom

Deep learning has been tremendously successful, especially for solving problems in natural language understanding and computer vision. For example, today’s best deep convolutional neural networks now rival humans at image recognition. This is a great achievement, but artificial intelligence still has a long way to go toward achieving the versatility of humans. In this talk, I describe two efforts by my lab to create deep learning algorithms that are more flexible and learn more like people. In the first part of the talk, I discuss my lab’s work developing Visual Question Answering (VQA) systems. In VQA, an algorithm must answer text-based queries about images. While performance on VQA benchmarks is rapidly approaching human-level, I argue we still have a long way to go. In the second part of the talk, I describe algorithms for lifelong learning in neural networks, which can learn over time without catastrophic forgetting. Lifelong machine learning has many applications, including robotics and learning on resource constrained embedded platforms. I conclude the talk with a discussion of efforts to promote and grow AI education and research at the Rochester Institute of Technology.

Christopher Kanan is an assistant professor in the Chester F. Carlson Center for Imaging Science at the Rochester Institute of Technology (RIT). At RIT, he is also Associate Director of the Center for Human-aware AI, a Computer Science Department affiliate faculty member, and a McNair Scholars advisory board member. Dr. Kanan’s lab uses deep learning to solve problems in AI, with an emphasis on lifelong machine learning and task-driven algorithms for understanding scenes. Dr. Kanan received a PhD in computer science from the University of California at San Diego, where he worked on brain-inspired algorithms for object recognition, neural networks, active vision, and cognitive modeling. He received an MS in computer science from the University of Southern California. Before joining RIT, Dr. Kanan was a postdoctoral scholar at the California Institute of Technology, and later worked as a Research Technologist at NASA’s Jet Propulsion Laboratory, where he used deep learning to develop vision systems for autonomous ships. He was the recipient of the 2016 Rising Star award in RIT's College of Science and is an IEEE Senior Member.

Email [email protected] to RSVP