Course: Using machine learning for dynamic data

« Back
Course title Using machine learning for dynamic data
Course code KEP/VSU
Organizational form of instruction Lecture + Tutorial
Level of course unspecified
Year of study not specified
Semester Summer
Number of ECTS credits 4
Language of instruction Czech, English
Status of course unspecified
Form of instruction Face-to-face
Work placements This is not an internship
Recommended optional programme components None
Lecturer(s)
  • Šmídl Václav, prof. Ing. Ph.D.
Course content
1. Definition of standard tasks in time series (prediction, fault detection, smoothing). 2. Classical approaches (fft, filtration). 3. Modern methods of time series prediction - recurrent neural networks. 4. Structured state models (SSM) transformers and their connections. 5. Processing data from multiple modalities (sensor fusion). 6. Applications in fault and anomaly detection. 7. Graph theory, graph neural networks. 8. Use of graph neural networks in electrical circuits. 9. Use of neural networks for solving and identifying differential equations (PINN). 10. Estimation of model prediction uncertainty in general (MC dropout, ensembles) and for individual special cases. 11. Use in active learning methods and device design. 12. Methods of model explainability (attention, attribution, abductive explanation).

Learning activities and teaching methods
One-to-One tutorial, Group discussion, Individual study
  • Individual project (40) - 40 hours per semester
  • Graduate study programme term essay (40-50) - 20 hours per semester
  • Contact hours - 52 hours per semester
prerequisite
Knowledge
to have basic knowledge of programming concepts in Python
to have basic knowledge of machine learning concepts (cross-validation, optimization, plausibility)
Skills
to operate commonly available computer technology
to use basic algorithms, data and control structures
Competences
N/A
N/A
N/A
learning outcomes
Knowledge
to explain modern machine learning algorithms and apply them to practical problems
to have knowledge of various types of data including time series, graphs, and differential equations
Skills
to apply modern machine learning algorithms to practical problems
to work with different types of data including time series, graphs, and differential equations
to apply methods to data from real-world problems
Competences
N/A
N/A
N/A
teaching methods
Knowledge
Lecture
Lecture supplemented with a discussion
Multimedia supported teaching
Skills
Practicum
Project-based instruction
Students' portfolio
Competences
Lecture supplemented with a discussion
Students' portfolio
assessment methods
Knowledge
Project
Continuous assessment
Skills
Individual presentation at a seminar
Skills demonstration during practicum
Competences
Individual presentation at a seminar
Continuous assessment
Recommended literature
  • Albert Gu, Karan Goel, Christopher Ré. Efficiently Modeling Long Sequences with Structured State Spaces. .
  • Goodfellow, I., Bengio, Y., Courville, A., & Bengio, Y. Deep learning. Cambridge. 2016.
  • Raissi, Maziar, Paris Perdikaris, and George E. Karniadakis. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. .
  • Zhang, Aston; Lipton, Zachary C.,; Li, Mu; Smola, Alexander J. Dive into deep learning. First published. 2024. ISBN 978-1-009-38943-3.


Study plans that include the course
Faculty Study plan (Version) Category of Branch/Specialization Recommended year of study Recommended semester