Modulcode:	Inf-MaLearn
Englische Bezeichnung:	Machine Learning
Modulverantwortliche(r):	Prof. Dr. Carsten Meyer
Turnus:	jedes Jahr im SS (SS14 SS15 SS16 SS17)
Präsenzzeiten:	2V 2Ü
ECTS:	6
Workload:	30 Std. Vorlesung, 30 Std. Präsenzübung, 120 Std. Selbststudium
Dauer:	ein Semester
Modulkategorien:	WI (BSc Inf (15)) PI (MSc Inf (15)) WI (MSc Inf (15)) WI (MEd Inf) WPI (MEd Inf) IG (TA) (MSc Inf (2-Fach)) IS (SA) (MSc Inf (2-Fach)) IG (SA) (MSc Inf (2-Fach)) IG (MSc Inf) IS (MSc Inf)
Lehrsprache:	Englisch
Voraussetzungen:

Kurzfassung:

Machine learning (a branch of artificial intelligence) is concerned with the design and development of algorithms that allow technical systems to solve tasks and to improve their performance by ("clever") learning from examples. The aim of this course is to provide a fundamental understanding of important concepts in machine learning, both from a theoretical and an application point of view. Several learning tasks (classification, regression, clustering), learning modes (supervised and unsupervised learning) and learning machines (support vector machine, multi-layer perceptron, decision tree) are covered, in addition to methods for dimensionality reduction (principal component analysis, linear discriminant analysis) and algorithms for model selection and model combination (bagging, boosting). The lecture also contains a brief introduction to deep learning.

Lernziele:

The students learn about fundamental concepts, algorithms and approaches in machine learning (including a short introduction to deep learning), typical problems and applications suited for machine learning as well as advantages and limitations of the individual methods. The concepts are introduced from a theoretical point of view in the lecture; the exercises shall deepen the theoretical understanding as well as present practical applications of the algorithms. To this end, available software libraries (based on the python programming language) are used to solve and analyze simple "toy" problems as well as real-world machine learning problems. Thus, after successful participation in the course, given a suitable real-world machine learning problem the students should be able to identify promising algorithms to solve the problem, to practically apply the algorithms to the problem using software tools, and to analyse and potentially improve the performance of the algorithm.

Lehrinhalte:

The following aspects will be covered in the lecture:

• Introduction, Machine learning basics
• Supervised classification: Support vector machines, decision trees, multi-layer perceptrons
• Unsupervised learning / clustering
• Dimensionality reduction: Linear discriminant analysis, principal component analysis
• Model selection
• Ensemble methods: Bagging, boosting
• Introduction to deep learning

The exercises contain theoretical and practical exercises (based on available software libraries written in the python programming language) to deepen the understanding of the algorithms.

Weitere Voraussetzungen:

Mathematical basics of algebra and analysis and of optimization.

Prüfungsleistung:

Exam at the end of the course (written exam if 6 participants or more, otherwise oral exam)

Lehr- und Lernmethoden:

Beamer presentation, use of software tools.

Verwendbarkeit:

Literatur:

• T. Mitchell, "Machine learning", McGraw Hill, 1997
• E. Alpaydin, "Introduction to Machine Learning", MIT Press, 2010
• S. Marsland, "Machine Learning: An Algorithmic Perspective", CRC Press, 2009
• C. M. Bishop, "Pattern recognition and Machine learning", Springer, 2006
• R. Duda et al., "Pattern classification", Wiley, 2001
• S. Haykin, "Neural networks and learning machines", Prentice Hall, 2008
• M. Nielsen, "Neural Networks and Deep Learning", http://neuralnetworksanddeeplearning.com/index.html

Verweise:

Kommentar:

Students are encouraged to bring their own laptops to the laboratory exercises; knowledge of the python programming language is not a prerequisite. Exercises are encouraged to be solved in teams of two persons.