Hey Prague, MLMU 2020 is here! And this February, we are going to focus on Deep Learning. Jiri Materna, Adam Kolar and Dusan Fedorcak will lead you through the foundations of deep learning in three fields: Natural Language Processing, Image Processing and Time Series Analysis.
Deep Learning in a Nutshell
This series of 3 deep learning talks is a short demo of our intensive machine learning courses for public taught at www.mlcollege.com. The talks are intended for beginners who have no or very limited experience with deep learning.
Please register here by clicking RSVP Yes! The number of seats is limited. (Entry is free)
18:00-18:30 Arrival of attendees
18:30-19:00 Jiri Materna: Natural Language Processing
19:00-19:30 Adam Kolar: Image Processing
19:30-20:00 Dusan Fedorcak: Time Series Analysis
Part 1: Natural Language Processing
The field of natural language processing has undergone many big changes during the past years. In this introductory talk we will briefly discuss what the biggest challenges in natural language processing are, and then dive into an overview of the most important deep learning milestones in NLP. We will namely cover word embeddings, recurrent neural networks for language modeling and machine translation, and the recent boom of Transformer-based models.
Part 2: Image Processing
In the session dedicated to deep learning over images, we will try to briefly address a few major breakthroughs of the last decade in the field covering topics like image classification, object detection or generative modeling on the image domain. How did the state of the art change over the years, which of the methods are adopted in the industry and what are still the challenges and open problems? This quick tour could help you to make a basic overview.
Part 3: Time Series Analysis
The third part of the talk will be dedicated to the time series analysis using deep learning. First, we look at the nature of temporal data and common tasks where machine learning is applied. Then, we'll briefly go through classical time series analysis to set a baseline for further discussion. After that, we dive into modern recurrent neural networks, where we discuss their purpose, advantages over past models and we'll look at how temporal data is prepared for these kind of neural networks. Finally, we'll show a couple examples where deep learning models are combined together for more robust time series prediction.
Jiří Materna: He is a machine learning expert with machine learning experience in industry since 2007. After finishing his Ph.D., he was working as the head of research at Seznam.cz and now offers machine learning solutions and consulting as a freelancer. He is the founder and lecturer at Machine Learning College and the organizer of an international conference Machine Learning Prague.
Dusan Fedorcak: During his doctoral studies, Dusan focused on self-organization networks, unsupervised learning, time series prediction in hydrology and traffic modeling. Since 2014, he has been involved in several start-up projects (GoodAI – general artificial intelligence, Neuron Soundware – deep learning for sound records in industry, CEAi – fintech & natural language processing).
Adam Kolar: After graduating from Brno University of Technology, he spent the biggest part of his professional career as a researcher and then the chief of one of the research teams at Seznam.cz. Now he is helping to set up machine learning startups for Central Europe AI and organizes Machine Learning Meetups Brno.