Satellite event takes place on 01, 08 & 15, July, 2018. These are events that may be of broad interest to the participants of the school. Events could be half day (3-4 Hrs) or full day (6-8 Hrs). Summer school organizers primarily facilitate this.
|Time||Sunday 08th July, 2018|
|08:00 - 09:00||Breakfast|
|09:00 - 13:00||Deep Learning for Computer Vision: Hands-on MATLAB Workshop|
|13:00 - 13:30||Lunch|
|13:30 - 16:30||AI on IA (Artificial Intelligence on Intel Architecture)|
|17:00 - 19:00||Fundamentals of Probabilistic graphical models|
Abstract : In computer vision and machine learning, several tasks involve random variables and often multiple of them. Generally, these random variables are dependent on each other in complex ways. Probabilistic graphical models are a popular framework for representing such dependencies among the random variables. Such graphical models have particularly proven to be versatile frameworks for encoding prior knowledge within machine learning models. Recently in context of deep learning, probabilistic graphical models have been popularly used in generative models. In this short tutorial, we shall discuss the basic principles associated with probabilistic graphical models, both directed graphical models (Bayesian networks) and undirected graphical models (Markov networks).
Abstract : Designing and deploying deep learning based computer vision applications to embedded CPU and GPU platforms is challenging because of resource constraints inherent in embedded devices. A MATLAB® based workflow facilitates the design of these applications, and automatically generated C or CUDA® code can be deployed on boards like the Jetson TX2 and DRIVE™ PX to achieve very fast inference. The workshop illustrates how MATLAB supports all major phases of this workflow. Starting with algorithm design, the algorithm may employ deep neural networks augmented with traditional computer vision techniques and can be tested and verified within MATLAB. Next, these networks are trained using GPU and parallel computing support for MATLAB either on the desktop, cluster, or the cloud. Finally, GPU Coder™ generates portable and optimized C/C++ and/or CUDA® code from the MATLAB algorithm, which is then cross-compiled and deployed to ARM/Intel CPUs and NVIDIA Tegra® boards.
|Time||Monday 09th July, 2018|
|18:00 - 20:00||Linear Algebra For ML|
Abstract : Linear Algebra is one of the foundation fields of mathematics that gives the essential tools to appreciate machine learning techniques. The formalisms used in linear algebra are used by other branches of mathematics to express the concepts relevant to this area. In fact, it is the bed rock of multi-variable calculus and statistics, two other areas needed to understand majority of literature in machine learning. This event would focus on two most important aspects related to matrices - their multiplication and factorisation. The first half would focus on building a geometrical intuition about these matrix operations. The second half would cover the applications of the concepts we discussed in the first part. The applications may include avoiding underflow/overflow when operating on matrices, formatting a real-world computer vision problem as system of linear equations and trying to solve them efficiently. Note that it is not a cover-all crash course but a very gentle introduction to thinking in terms of matrices. This will be helpful in case you are uncomfortable with matrix notation/operations or you want to have deeper intuition for vectors and matrices.