Title: Deep Learning
Deep learning methods, a reincarnation of neural networks, have demonstrated impressive performance in a variety of applications domains in recent years including computer vision, natural language processing and speech understanding. Silently, but surely, these methods have become a part of every technology consumer’s life through their use in applications such as image search, face detection and speech processing. It is widely understood now that deep learning methods use deep graph architectures to learn hierarchical representations of data using linear and non-linear transformations, which are then used for learning and inference. Deep learning architectures include Convolutional Neural Networks, Recurrent Neural Networks, Autoencoders, Boltzmann Machines, and more recently, improvised networks such as Residual Networks and Memory Networks. These various morphologies of deep learning have displayed immense potential in solving AI problems that have challenged the field over the past few decades, supported by the availability of large amounts of data and computational power. This 3-hour tutorial will provide an introduction to deep learning, its mathematical foundations, a description of the different architectures, their applications for multimedia processing, especially in computer vision, and some recent trends.
Vineeth N Balasubramanian is an Assistant Professor in the Department of Computer Science and Engineering at the Indian Institute of Technology, Hyderabad, India. His research interests include machine learning, computer vision and their applications to real-world problems, broadly in human behavior understanding. He completed his PhD at Arizona State University in 2010, where his dissertation on the Conformal Prediction framework was nominated for the Outstanding PhD Dissertation at the Department of Computer Science. He has over 50 research publications including articles in premier peer-reviewed venues (such as IEEE TPAMI, IEEE TNNLS, ACM KDD, IEEE CVPR, IEEE ICDM), 3 patents under review and an edited book on a recent development in machine learning called Conformal Prediction. He has also received Best Paper Awards, including at the IEEE HiPC Student Research Symposium 2014, as well as at IEEE HAVE 2008. His current research focuses on deep learning, its applications and its interconnection with optimization methods.