Predicting the dynamical behaviour in complex problems with machine learning: GNS.

Engineering disciple stems from the ingenuity and the ability to contrive. This heritage has also been reflected on how the raw data is processed into information and distilled into engineering wisdom. Since its dawn, engineers have been dealing with massive amounts of data accumulated over decades of fundamental experiments and field measurements, vitalized in the form of cleverly organized charts, tables and heuristic laws. In the last few decades, our capability to generate data has increased even further with the developments in (i) the measurement techniques including the sensing technologies, (ii) computational power, (iii) faster, easier and cheaper data transfer and storage, (iv) post-processing tools and algorithms. In this regard, machine learning (ML) can be interpreted as a powerful tool that augments the traditional "wisdom distillation process" via automatic identification of the patterns hidden within the data.

This course focuses on the fundamentals of intelligence and ML methodologies that can be utilized to solve engineering problems. Throughout the semester, students will build up the basic skills needed to develop intelligent solutions for pattern recognition in experimental/numerical datasets, model abstractions, optimization and process control. Another objective is to establish a strong scientific background in order to clarify the current capabilities, challenges and opportunities in ML. The course includes weekly software sessions in TensorFlow for hands-on experience and finalizes with an end-to-end Machine Learning Project.

Content

1. Introduction to Data Driven Engineering
2. Basics of Learning
3. Analysis of Static Datasets I: Classification and Regression
4. Analysis of Static Datasets II: Clustering and Dimensionality Reduction
5. Deep Learning for Dynamic Systems
6. Sequence Modeling
7. Generative Modeling
8. Machine Learning Control
9. Emerging Concepts and the Outlook
10. Project Sessions

Lecture Format

• Lectures: 45 min; Practice hours: 45 min
• Course capacity: Limited

Workload

• Regular attendance: 21 hours
• self-study: 42 hours

Learning Objectives

Students have the ability to:

• distinguish between different learning methods (information, similarity, probability, error-based) and select right strategies and algorithms accordingly,
• explore large datasets, handle data quality issues and prepare it for downstream applications,
• explain the procedures of ML algorithms,
• judge and apply different approaches to analyze static datasets,
• analyze and evaluate methods for large dynamic systems via deep learning,
• plan and execute an end-to-end machine learning project from data collection to launch phase,
• design recipes for a given problem and to solve practical engineering problems with ML.

Recommendations:

The course requires basic knowledge in engineering mathematics and computer programming at an undergraduate level. It is strongly recommended to be taken in combination with the lecture "Data Driven Engineering 2: Advanced Topics".