From Machine Learning to Deep Learning: A concise introduction

Prerequisites

• Familiarity with Linux operating systems, including Linux shell (some parts of the training will use a cluster).
• Technical background and basic understanding of machine learning concepts will be helpful.
• Preliminary experience with Python is required. Since Python is used, the following tutorial can be used to learn the syntax.
• For the second day, familiarity with TensorFlow will be a plus as all hands-on sessions will be using TensorFlow. For those who like to use TensorFlow in advance TensorFlow tutorial will be a great help.

Content levels

• Basic: 10:30 hours
• Intermediate: 2:30 hours
• Advanced: 1 hour
• Community: 6:30 hours

Learn more about course curricula and content levels.

After this course, participants will

• have a basic understanding of classical Machine Learning and Deep Learning (DL) concepts and methods,
• have gained practical experience in applying these methods,
• and will know how to use HLRS's systems for certain ML or DL tasks.

Nico Formanek, Dr. Khatuna Kakhiani, Patrick Vogler and Dr.-Ing. Lorenzo Zanon (HLRS), and Anna Schwarz (IAG).

(preliminary)

08:30 - 09:00  Local registration

Day 1: Focus on Pre-processing, Feature Engineering and Machine Learning (9:00 - 17:30, Dr.-Ing. Lorenzo Zanon)

The first day will be based on the “Stuttgart S-Bahn Example” (originally developed by Dennis Hoppe, HLRS) to provide an introduction to Machine Learning. The focus is on data preparation, classification and regression algorithms in supervised learning: Can these tools be helpful to improve the travel experience in the Stuttgart S-Bahn, which are their limits? Apache Spark will be employed for the hands-on sessions on Jupyter Notebooks as well as via interactive jobs on script. Finally, we will also touch upon the visualisation of results.

Day 2: Focus on supervised Deep Learning to classify images of waste in the wild (9:00 - 17:30, Dr. Khatuna Kakhiani)

During this day, participants will explore how Deep Learning can be used to classificy waste in wild. After brief introduction of Deep Learning, and basic concepts and Building blocks of Deep Neural Networks, participants will learn how to:

• implement common deep learning workflow for classification problem
• preprocess unstructured data, understand image data, manipulate image instances and visualize results, work with unbalanced small dataset
• define the Neural Network Architecture (Sigmoid Neural Model, Multilayer Perceptron (MLP), Convolutional Neural Networks (CNN)); Compile, Train and Evaluate the Model
• classify images with Convolutional Neural Networks (CNN)
• visualize results

Upon completion, participant should be able to solve classification problems with CNN on other custom datasets. Tensorflow will be employed for the hands-on sessions on Jupyter Notebooks.

Day 3:

• Guest Lecture: Towards Data-Driven Computational Fluid Dynamics (9:00 - 11:30, Anna Schwarz, IAG)
• Generalization and the problem of leakage: (11:45-13:00, Nico Formanek, HLRS)
• Lunch break: 13:00-14:15
• Data Compression of numerical data sets with the BigWhoop library (14:15-16:45, Patrick Vogler, HLRS)

On the third day we start with the guest lecture "Towards Data-Driven Computational Fluid Dynamics". It will be given by Anna Schwarz, Institute of Aerodynamics and Gas Dynamics, University of Stuttgart.

We will conclude the day with an introduction to data compression, focusing on the various methods available to us for the efficient size reduction of our training data. Special attention will be paid to which approaches are best suited for different data types and what impact the different approaches and compression rates have on the quality of the datasets. The compression library BigWhoop and its accompanying command line tool will be made available for the hands-on sessions.

Lunch break will be from 13:00-14:15.

The exercises will be carried out on HLRS's systems using Jupyter Notebooks.

Registration closes on March 9, 2024.

Late registrations after that date might still be possible according to the course capacity.

• Students without master’s degree or equivalent: 30 EUR
• PhD students or employees at a German university or public research institute: 60 EUR
• PhD students or employees at a university or public research institute in an EU, EU-associated or PRACE country other than Germany: 120 EUR.
• PhD students or employees at a university or public research institute outside of EU, EU-associated or PRACE countries: 240 EUR
• Other participants, e.g., from industry, other public service providers, or government: 600 EUR

Link to the EU and EU-associated (Horizon Europe), and PRACE countries.

Our course fee includes coffee breaks (in classroom courses only).

This course is partly realised in cooperation with the Centre of Excellence EXCELLERAT P2. Funded by the European Union. This work has received funding from the European High Performance Computing Joint Undertaking (JU) and Germany, Italy, Slovenia, Spain, Sweden, and France under grant agreement No 101092621. See also the EXCELLERAT Service Portal for more information.

This course is partly realised in cooperation with the Centre of Excellence CEEC. Funded by the European Union. This work has received funding from the European High Performance Computing Joint Undertaking (JU) and Sweden, Germany, Spain, Greece, and Denmark under grant agreement No 101093393.

Tobias Haas, phone 0711 685 87223, tobias.haas(at)hlrs.de