BMVC 2013 - Monday Tutorials
TUTORIAL on Deep Learning for Machine Vision
by Adam Coates (Stanford)
Abstract. Machine learning algorithms have freed practitioners from many error-prone, hand-engineered components for making decisions in common machine vision tasks such as object recognition. A major source of difficulty, however, is that such learning systems still rely on many hand-built components like sophisticated feature extractors that attempt to identify higher-level patterns in images that typical learning algorithms cannot discover on their own. "Deep learning" and "representation learning" algorithms aim to remove this hurdle by learning higher-level representations automatically from data and have led to recent successes in vision, speech, and language tasks. This tutorial will introduce the basic components of deep learning algorithms and practical techniques for debugging and applying these methods to machine vision problems. The first part of the tutorial will cover neural network models and basic training approaches including error back-propagation and numerical optimization methods, with image classification as a motivating application. The second part will cover additional (sometimes domain-specific) techniques to improve the performance of these algorithms and apply them to other vision tasks including detection and image segmentation. With these tools, audience members will understand how deep learning algorithms work and how they are used in practical applications with sufficient knowledge to complete a hands-on tutorial available on the web. We will conclude with a brief high-level overview of other important topics and results in deep learning research.
Biography. Dr Adam Coates received his PhD in Computer Science from Stanford University in 2012. He is currently a post-doctoral researcher at Stanford and a Visiting Scholar at Indiana University, Bloomington. His research focuses on scaling up machine learning and representation learning algorithms to enable machines to acquire knowledge from unsupervised experience. His interests cover related key topics in computer vision, reinforcement learning, and robotics. He has received Best Student Paper awards from ICML and ICDAR; but his favorite endorsement is still the Instrument Rating on his pilot's license.
TUTORIAL on Submodularity in Machine Learning and Vision
by Andreas Krause (ETH Zurich)
Abstract. Numerous problems in machine learning and vision are inherently discrete. More often than not, these lead to challenging optimization problems. While convexity is an important property when solving continuous optimization problems, submodularity, often viewed as a discrete analog of convexity, is key to solving many discrete problems. Its characterizing property, diminishing marginal returns, appears naturally in a multitude of settings. While submodularity has long been recognized in combinatorial optimization and game theory, it has seen a recent surge of interest in theoretical computer science, machine learning and computer vision. This tutorial will introduce the concept of submodularity and its basic properties, and outline recent research directions -- such as new approaches towards large-scale optimization and sequential decision making tasks. We will discuss recent applications to challenging machine learning and vision problems such as high-order graphical model inference, structured sparse modeling, multiple object detection, active sensing etc. The tutorial will not assume any specific prior knowledge on the subject.
Biography. Andreas Krause is an Assistant Professor of Computer Science at ETH Zurich, where he leads the Learning & Adaptive Systems Group. Before that he was Assistant Professor of Computer Science at Caltech (2009-2012). He received his Ph.D. and M.Sc. in Computer Science from Carnegie Mellon University (2008) and his Diplom in Computer Science and Mathematics from the Technical University of Munich, Germany (2004). He is a Microsoft Research Faculty Fellow and a Kavli Frontiers Fellow of the US National Academy of Sciences. He received an ERC Starting Investigator grant, the Deutscher Mustererkennungspreis, an NSF CAREER award, the Okawa Foundation Research Grant recognizing top young researchers in telecommunications as well as the ETH Golden Owl teaching award. His research in learning and adaptive systems that actively acquire information, reason and make decisions in large, distributed and uncertain domains received awards at premier conferences (AAAI, KDD, IPSN, ICML, UAI) and journals (JAIR, JWRPM).