Special lecture

Introduction to Machine Learning
Prof. Dr. Martin Riedmiller, Dr. Frank Hutter, Manuel Blum

• Dates and Place (Vorlesungsverzeichnis):
  • Tuesday, 16:00 - 17:30, Geb. 51, Raum 00-031
  • Donnerstag, 8:15 - 9:45, Geb. 51, Raum 00-031
• Exercises (Vorlesungsverzeichnis):
  • Exercise dates: 20.5., 22.5., 24.6., 26.6., 15.7., 17.7., 31.7. presentation
• Credits:
  • 6 ECTS
• Exam:
  • registration via examination office
  • this semester, we have an increased focus on practical ML experience, which is also relevant for the exam. It is therefore highly recommended to participate in the practical exercises (see also 'Organisation of Exercises' below).
  • all master students and bachelor ESE: written exam. date: 25.09.2014, 9:00, Geb. 101, 01-009/13
  • bachelor students CS: oral exam in groups. date: 25.09.2014, Geb. 079, office Riedmiller
• Organisation issues
  • see also slides lecture 1 (introduction)
  • July, 22th: Lecture on RL
  • July, 24th: Exercise sheet 5
  • July, 29th: Lecture
  • July, 31th: Project presentations
• Exam preparation meeting on Friday, September 19th, 11:00, in 79/00019
  • you can ask questions about the exams and the exercises

• Organisation of Exercises

• Introduction
• Concept Learning
• Perceptrons
• MLPs
• Generalisation
• Boosting
• Decision Trees
  • Paper mentioned in class: J. Shotton et al., "Real-Time Human Pose Recognition in Parts from a Single Depth Image", CVPR 2011 (video)
  • Comprehensive recent overview about ensembles of decision trees: A. Criminisi et al., "Decision Forests for Classification, Regression, Density Estimation, Manifold Learning and Semi-Supervised Learning", Microsoft Technical Report MSR-TR-2011-114, 2011
• Hyperparameter Optimization
• Hyperparameter Optimization II
• Reinforcement Learning

• Key concepts (22.5.)
• Sheet 1 - Solution
• Sheet 2 - Solution
• Sheet 3 - Solution
• Sheet 4 - Solution
• Sheet 5 - Solution

• Iris demo iPython notebook: iris.ipynb
• recommended tools: tools.pdf
• Programming Task I
• Programming Task II

Please ask questions concerning HPOLib on google groups.

Die Vorlesung gibt eine Einführung in das Forschungsgebiet Maschinelles Lernen. Behandelt werden Methoden des überwachten, unüberwachten und optimierenden Lernens. Themengebiete sind unter anderem: Lernen im Hypothesenraum, Entscheidungsbäume, neuronale Netze…