|
Mathematical Methods for Machine Learning and Signal Processing (MMMLSP)
- Lecturer
- Prof. Dr. Veniamin Morgenshtern
- Details
- Vorlesung
4 cred.h, ECTS studies, ECTS credits: 5
nur Fachstudium, Sprache Englisch
Time and place: Tue 8:15 - 9:45, 05.025; Thu 12:15 - 13:45, 05.025
- Fields of study
- WPF ASC-MA ab 1 (ECTS-Credits: 5)
WF CME-MA ab 1 (ECTS-Credits: 5)
- Contents
- This course focuses on modern machine learning and signal processing algorithms that have firm mathematical footing.
First, we will study the basics of Frame Theory -- a mathematical framework for linear redundant signal expansions. We will discuss an applications in signal sampling.
Second, we will study the theory of Compressed Sensing -- a powerful way to recover sparse signals from an incomplete set of measurements.
Third, we will discuss applications of Compressed Sensing-based methods in Machine Learning: Matrix Completion and Subspace Clustering.
Finally, we will study the theory of Scattering Transform -- a signal representation based on deep neural network that is invariant to signal translations and deformations. This topic is one of the few mathematical results related to theoretical understanding of deep learning.
Time permitting, we will discuss other results related to the theoretical understanding of deep learning, such as Tishbi's Information Bottleneck principle.
- ECTS information:
- Credits: 5
- Contents
- This course focuses on modern machine learning and signal processing algorithms that have firm mathematical footing.
First, we will study the basics of Frame Theory -- a mathematical framework for linear redundant signal expansions. We will discuss an applications in signal sampling.
Second, we will study the theory of Compressed Sensing -- a powerful way to recover sparse signals from an incomplete set of measurements.
Third, we will discuss applications of Compressed Sensing-based methods in Machine Learning: Matrix Completion and Subspace Clustering.
Finally, we will study the theory of Scattering Transform -- a signal representation based on deep neural network that is invariant to signal translations and deformations. This topic is one of the few mathematical results related to theoretical understanding of deep learning.
Time permitting, we will discuss other results related to the theoretical understanding of deep learning, such as Tishbi's Information Bottleneck principle.
- Additional information
- Expected participants: 20, Maximale Teilnehmerzahl: 30
www: https://lms.tf.fau.de/studium-und-lehre/lehrveranstaltungen/
- Verwendung in folgenden UnivIS-Modulen
- Startsemester SS 2019:
- Mathematical Methods for Machine Learning and Signal Processing (MMMLSP)
- Department: Chair of Multimedia Communications and Signal Processing (Prof. Dr. Kaup)
|
|
|