Thomas Demeester (Ghent University) (2017-03-17 10:45 - 15:30 in ZI-2126)
Thomas Demeester will give a half-day tutorial on TensorFlow. TensorFlow was developed by Google, with the goal of “bringing machine learning to everyone”, with the focus on deep neural networks. This open source library allows to specify high-level computations as data flow graphs. Nodes in the graph represent mathematical operations, while the graph edges represent the multidimensional data arrays (tensors) communicated between them. The architecture allows you to efficiently deploy computation to one or more CPUs or GPUs in a desktop or cluster with a single API. The tutorial is meant for people who have some notions of machine learning, no experience whatsoever with TensorFlow, and want to get their hands dirty. The goal is to tackle the initial burden of an unknown framework for deep learning, to give people understanding and experience of building basic deep learning models. The requirement for anyone attending would be to have a laptop with TensorFlow and Jupyter installed (instructions will be sent after registering below).
Please register via: https://registration.closed
We have room for only a limited number of participants.
Thomas Demeester is a post-doctoral researcher at Ghent University - imec, Belgium. He holds a PhD in electrical engineering, for which he modeled electromagnetic phenomena on interconnections. After his PhD, Demeester worked on several natural language processing and information retrieval problems including federated search (in collaboration with the University of Twente Database Group), user modeling, knowledge base population, and keyphrase extraction. Recently, as a visiting researcher at University College London, Demeester worked on neural link prediction and sequence encoding with deep neural networks.