Moses - Machine Intelligence

Display language

Diese Modulbeschreibung gilt nicht im aktuellen Semester.

Die Modulbeschreibung gilt nur für den Zeitraum SS 2017 - SoSe 2021. Die Version für das aktuelle Semester (SoSe 2024) finden Sie hier: Modulbeschreibung des SoSe 2024

Module / Version:
#40786 / #3

Validity:
SS 2017 - SoSe 2021

Default display language:
English

Module title:
Machine Intelligence

Credits:
12

Responsible person:
Obermayer, Klaus

Grading:
benotet

Type of exam:
Mündliche Prüfung

Zugehörigkeit

Faculty:
Fakultät IV

Institute:
Institut für Softwaretechnik und Theoretische Informatik

Area of expertise:
34351300 FG Neuronale Informationsverarbeitung

Prüfungsausschuss:
No information

Kontakt

Office:
MAR 5-6

Contact person:
Groiß, Camilla

E-mail address:
sekr@ni.tu-berlin.de

Website:

https://www.bccn-berlin.de/courses-and-modules.html

Learning Outcomes

In this module, participants will gain knowledge about: - basic concepts, their theoretical foundation and the most common algorithms used in machine learning and artificial intelligence - strengths and limitations of the different paradigms They will be enabled to: - apply methods and algorithms to real world problems - be aware of performance criteria - critically evaluate results obtained with those methods - modify algorithms to new tasks at hand - develop new algorithms according to the paradigms presented in this course.

Content

Part 1: Artificial neural networks. Connectionist neurons, the multilayer perceptron, radial basis function networks, learning by empirical risk minimization, gradient-based optimization, overfitting and underfitting, regularization techniques, deep networks, applications to classification and regression problems. Part 2: Learning theory and support vector machines. Elements of statistical learning theory, learning by structural risk minimization, the C Support Vector Machine, kernels and non-linear decision boundaries, SMO optimization, the P-SVM. Part 3: Probabilistic methods. Reasoning under uncertainty and Bayesian inference; graphical models, graphs vs. distributions, and belief propagation; generative models; Bayesian inference and neural networks; non-parametric density estimation; parametric density estimation and maximum likelihood methods. Part 4: Reinforcement learning (MDP, value iteration, policy iteration, Q-learning). Part 5: Projections methods. Principal Component Analysis and Kernel-PCA; independent component analysis and blind source separation techniques (Infomax, Fast-ICA, ESD). Part 6: Stochastic optimization. Simulated annealing, mean-field techniques. Part 7: Clustering and embedding. K-means clustering, pairwise clustering methods, self-organizing maps for central and pairwise data.

Module Components

Pflichtgruppe:

All Courses are mandatory.

Course Name	Type	Number	Cycle	Language	SWS
Machine Intelligence I	VL	0434 L 866	WiSe	No information	2
Machine Intelligence II	VL	0434 L 867	SoSe	No information	2
Machine Intelligence I	UE	0434 L 866	WiSe	No information	2
Machine Intelligence II	UE	0434 L 867	SoSe	No information	2

Workload and Credit Points

Machine Intelligence I (VL):

Workload description	Multiplier	Hours	Total
Präsenzzeit	15.0	2.0h	30.0h
Vor-/Nachbereitung	15.0	2.0h	30.0h
			60.0h(~2 LP)

Machine Intelligence II (VL):

Workload description	Multiplier	Hours	Total
Präsenzzeit	15.0	2.0h	30.0h
Vor-/Nachbereitung	15.0	2.0h	30.0h
			60.0h(~2 LP)

Machine Intelligence I (UE):

Workload description	Multiplier	Hours	Total
Präsenzzeit	15.0	2.0h	30.0h
Vor-/Nachbereitung	15.0	6.0h	90.0h
			120.0h(~4 LP)

Machine Intelligence II (UE):

Workload description	Multiplier	Hours	Total
Präsenzzeit	15.0	2.0h	30.0h
Vor-/Nachbereitung	15.0	6.0h	90.0h
			120.0h(~4 LP)

The Workload of the module sums up to 360.0 Hours. Therefore the module contains 12 Credits.

Description of Teaching and Learning Methods

The lecture part consists of teaching in front of the class. Participants are expected to rehearse topics after class, using their class notes as well as recommended book chapters, in preparation for the exercises and tutorials. Homework assignments are given on a regular basis, and must be usually solved within one or two weeks. These assignments cover analytical & mathematical exercises as well as numerical simulations & programming exercises. Working in small groups of two to three students is encouraged. Homework assignments and their solutions are discussed during the tutorial. In addition, selected topics presented during the lecture are rehearsed by the tutor as needed. The first tutorials cover a brief mathematics primer, and recommendations are provided for students for the module “individual studies”, if deficits in their mathematical knowledge become obvious.

Requirements for participation and examination

Desirable prerequisites for participation in the courses:

Wünschenswerte Voraussetzungen für die Teilnahme zu den Lehrveranstaltungen: Mathematical knowledge: Analysis, linear algebra, probability calculus and statistics, on a level comparable to mathematics courses for engineers (worth 24 credit points). Basic programming skills. Good command of the English language.

Mandatory requirements for the module test application:

1. Requirement

[NI] Machine Intelligence II - Hausaufgabe

2. Requirement

[NI] Machine Intelligence I - Hausaufgabe

Module completion

Grading

graded

Type of exam

Oral exam

Language

English

Duration/Extent

30 Min.

Duration of the Module

The following number of semesters is estimated for taking and completing the module:
2 Semester.

This module may be commenced in the following semesters:
Winter- und Sommersemester.

Maximum Number of Participants

This module is not limited to a number of students.

Registration Procedures

Enrollment to the module is handled in the first class of each module component (cf. 3). Students must be present in person. The module components Machine Intelligence I (lecture with exercises) and Machine Intelligence II (lecture with exercises) can be taken in any order, i.e. students may also start the module in the summer term. To be allowed to do the oral exam, students must achieve (seperately) at least 60% of the points awarded for homework in each of the two lectures. Students of the Master program in Computational Neuroscience have to register for the final oral exam at least three working days prior to the examination date. Registration has to be done with the examination office (Prüfungsamt) of TU Berlin. For students from other programs, other regulations may apply. Please consult the examination regulations (Prüfungsordnung) of your program. sekr@ni.tu-berlin.de

Recommended literature
01. Bishop, Pattern Recognition and Machine Learning, Springer-Verlag, 2006. (recommended)
02. Duda, Hart, Stock, Pattern Classification, Wiley, 2000. (recommended)
03. Haykin, Neural Networks, Prentice Hall, 1998. (recommended)
04. Kohonen, Self-Organizing Maps, Springer-Verlag, 1997. (recommended)
05. Schölkopf, Smola, Learning with Kernels, MIT Press, 2002. (recommended)
06. Russel, Norvig, Artificial Intelligence, Prentice Hall, 2003, Chapters 13 and 14. (recommended)
07. Cichocki, Amari, Adaptive Blind Signal and Image Processing, Wiley, 2002. (additional)
08. Cowell, Dawid, Lauritzen, Spiegelhalter, Probabilistic Networks and Expert Systems, Springer Verlag, 1999. (additional)
09. Hyvärinen, Karhunen, Oja, Independent Component Analysis, Wiley, 2001. (additional)
10. Jordan (Editor), Learning in Graphical Models, MIT Press, 1999. (additional)
11. Kay, Fundamentals of Statistical Signal Processing - Vol. I: Estimation Theory, Prentice Hall, 1993. (additional)
12. Ripley, Pattern Recognition and Neural Networks, Cambridge University Press, 1996. (additional)
13. Vapnik, Statistical Learning Theory, Wiley, 1998. (additional)
One or two specific book chapters are assigned / recomended to every topic of the lecture. This list of recommendations is explained during the first class of every module component and is available via TU Berlin’s ISIS platform

Assigned Degree Programs

Semester:

Zeige auch ausgelaufene Studiengänge und StuPOs

This module is used in the following Degree Programs (new System):

	Studiengang / StuPO	StuPOs	Verwendungen	Erste Verwendung	Letzte Verwendung
This module is not used in any degree program.

Miscellaneous

Das Modul ist exclusiv Studenten des Studiengangs „Computational Neuroscience (MSc) vorbehalten. Studenten anderer Studiengänge sollten statt dessen die beiden Module „Machine Intelligence 1“ und „Machine Intelligence 2“ belegen. The modul is restricted to students of „computational neuroscience (Msc)“. All other students should instead register for the courses „machine intelligence 1“ and „machine intelligence 2“.

Machine Intelligence