Jonathon Shlens
[introductory/intermediate] Learning a Representation of the Visual World with Neural Networks [virtual]
Summary
My lectures will provide a broad introduction to computer vision and convolutional neural networks. I will start by providing some historical context for whence these ideas have arrived and motivations for the systems we build today. I will move on to the basics of neural networks and convolutions and cover why and how we use these types of systems. This lecture will then move on to more modern considerations for building state of the art computer vision systems and provide a survey of common techniques in practice today. I will conclude by highlighting new directions for providing introspection for learned representations of vision as new as well as future directions for the field at large.
Syllabus
- History and motivations of computer vision
- Neural networks, deep learning and convolutions
- Modern advances in architectures
- Introspection of learned representations
- Opportunities and future directions
References
Goodfellow, Ian, Yoshua Bengio, and Aaron Courville. Deep learning. Cambridge, Mass: MIT Press, 2016. Chapter 9
Duda, Richard O., Peter E. Hart, and David G. Stork. Pattern classification and scene analysis. New York: Wiley, 2000
Pre-requisites
Linear algebra, Basic probability and statistics.
Short bio
I am a research scientist at Apple leading a machine learning research team dedicated to problems in vision, and basic science research. I recently moved from Google Brain to Apple. Back in the day, I received my Ph.D in computational neuroscience from UC San Diego where my research focused on applying machine learning towards understanding visual processing in real biological systems. My research interests have included the development of state-of-the-art computer vision systems, training algorithms for deep networks, generative models of images, and methods in computational neuroscience. You can see some of my publications at https://scholar.google.com/citations?&user=sm1q2bYAAAAJ
