Data Mining 2:
Artificial Neural Networks and Deep Learning

Lecture in Summer 2018

Overview

• Description
• Times and Rooms
• Lecturer
• Exam
• Lecture Slides
• Exercise Sheets
• Literature

Description

On this web page you can find information about the lecture "Data Mining 2: Artificial Neural Networks and Deep Learning" that is given by Christian Borgelt in summer 2018 at the University of Konstanz. This page will be updated in the course of the semester.

Type of activity:		Lecture/Exercise Lesson
Semester week hours:		2+2
Credits:		6
Language:		German or English according to preference
Requirements:		Data Mining 1 or Analysis and Visualization of Information
Certificate:		Written exam (2 hours), see Exam

Contents:

• Biological background
• Threshold logic units and their geometric interpretation
• General neural networks
• Multilayer perceptrons
• Deep learning
• Radial basis function networks
• Learning vector quantization and self-organizing maps
• Hopfield networks and associative memory
• Recurrent neural networks

Learning Objectives:

• Knowledge of the most common types of artificial neural networks
• Suitability of different network types for supervised and unsupervised learning tasks
• Understanding of the advantages and disadvantages of the different types of neural networks depending on the structure of the data
• Training neural networks with gradient descent and its variants
• Ability to select a neural network type for a given problem
• Application of training artificial neural networks in practice

back to the top

Times and Rooms

From the fourth semester week on, there will be two exercise lessons. The content of these exercise lessons will be the same; so only one of them needs to be attended.

back to the top

Lecturers

If you have questions about the lecture or the exercises, please refer (by email or in person) to:

• Lecture: Christian Borgelt, christian@borgelt.net, Room Z811
• Exercises: Christoph Doell, christoph.doell@uni-konstanz.de, Room Z812

There are no formal office hours. Simply check whether the office door is open!

back to the top

Exam

Exam Admission

The lecture is accompanied by exercise sheets. At the beginning of each exercise lesson, the exercises are voted for. By voting for an exercise, one expresses one's willingness to present something about it. (Suggestions for a solution will be discussed, they need not be correct right away.) To the exam will be admitted, who

• voted for at least half of the exercises and
• presented something for at least two exercises.

Exams

The first exam for the lecture will be a written exam, taking place on

Tuesday, 24.07.2018, 11:00 to 13:00 hours, in Room R511.

The second exam for the lecture will be a written exam, taking place on

to be determined.

back to the top

Lecture Slides

Lecture slides (in English, version 2019.04.24, 480 slides).

back to the top

Exercise Sheets

Here you can find all already prepared exercise sheets as well as the corresponding earliest exercise lesson date at which these exercise sheets will be discussed. However, the exercise sheets may, depending on how the exercise lessons proceed, be (partially) discussed later than the stated date.

• Exercise Sheet 01: nn01_eng.pdf nn01_eng.tex (from 2018.04.24)
• Exercise Sheet 02: nn02_eng.pdf nn02_eng.tex (from 2018.05.03)
• Exercise Sheet 03: nn03_eng.pdf nn03_eng.tex (from 2018.05.15)
• Exercise Sheet 04: nn04_eng.pdf nn04_eng.tex (from 2018.05.22)
• Exercise Sheet 05: nn05_eng.pdf nn05_eng.tex (from 2018.06.05)
• Exercise Sheet 06: nn06_eng.pdf nn06_eng.tex (from 2018.06.12)
• Exercise Sheet 07: nn07_eng.pdf nn07_eng.tex (from 2018.06.19)
• Exercise Sheet 08: nn08_eng.pdf nn08_eng.tex (from 2018.06.26)
• Exercise Sheet 09: nn09_eng.pdf nn09_eng.tex (from 2018.07.03)
• Exercise Sheet 10: nn10_eng.pdf nn10_eng.tex (from 2018.07.10)
• Exercise Sheet 11: nn11_eng.pdf nn11_eng.tex (from 2018.07.17)
• METAPOST sources for figures: nnex.mp graph.mp arrows.mp

Supplementary material:

• not yet available

back to the top

Literature

• Computational Intelligence - A Methodological Introduction
  Rudolf Kruse, Christian Borgelt, Christian Braune, Sanaz Mostaghim and Matthias Steinbrecher
  Springer-Verlag, Berlin/Heidelberg, Germany 2016, ISBN 978-1-4471-5012-1
  (2nd edition 2016, 564 pages, in English)
• Computational Intelligence - Eine methodische Einführung in Künstliche Neuronale Netze, Evolutionäre Algorithmen, Fuzzy-Systeme und Bayes-Netze
  Rudolf Kruse, Christian Borgelt, Christian Braune, Frank Klawonn, Christian Moewes, und Matthias Steinbrecher
  Springer-Vieweg, Berlin/Heidelberg, Germany 2015, ISBN 978-3-658-10904-2
  (2. Auflage 2015, 515 pages, in German)

back to the top