Navigation

Inhalt Hotkeys

Maschinelles Lernen

Machine learning

Vorlesung: Montag, 17:15 - 18:45, 1/305 (Dr. J. Vitay)
Übung: Mittwoch, 09:15 - 10:45, 1/B202 (R. Larisch)
Übung: Mittwoch, 15:30 - 17:00, 1/B202 (Dr. M. Teichmann)
Übung: Freitag, 11:30 - 13:00, 1/B202 (Dr. M. Teichmann)

General Information

Prerequisites: Modules in Mathematics I to IV, basic knowledge in Python.

Exam: written examination (90 minutes), 5 credit points.

Contact: julien dot vitay at informatik dot tu-chemnitz dot de.

Language: English. The exam can of course be done in German.

Exam WS2018-19

The written exam will be on Tuesday 19.02 at 13:30 in the room 1/316 only! The 1/201 will not be used.

Remarks and sample questions for the exam.

Content of the exam: everything except: 1) Restricted Boltzmann Machines (RBM), 2) Policy Gradient (deep RL), 3) AlphaGo.

Content

The course will present an introduction to the research field of Machine Learning, including Supervised Learning, Deep Learning and Reinforcement Learning. The different algorithms presented during the lectures will be studied in more details during the exercises, through implementations in Python.

The plan of the course is:

  1. Supervised learning
    1. Linear algorithms (regression, classification, softmax, maximum likelihood)
    2. Learning Theory (cross-validation, VC dimension, feature space)
    3. Neural Networks (MLP, regularization)
    4. Support vector machines (maximum margin classifier, kernel trick)
    5. Deep Learning (CNN, GAN)
    6. Recurrent neural networks (LSTM, GRU)
  2. Reinforcement Learning
    1. Formal definition of the RL-Problem (Markov Decision Processes)
    2. Dynamic Programming, Monte Carlo Methods
    3. Temporal Difference Learning (TD, Q-learning), Eligibility traces
    4. Deep Reinforcement learning (DQN, A3C, DDPG)

Literature

FAQ

  • How do I register for the course?
    You don't. There is no OPAL or anything, just show up to the lectures/exercises and register for the exam in December.
  • How do I register for the exam?
    search for Registration on SBService happens in December. Only registered students can participate to the exam.
  • I cannot assist to the exercises. Can I take the exam anyway?
    Yes. The exercises are there to help you understand the concepts seen in the lectures and get practical experience with machine learning tools. But they are not obligatory for the exam.
  • Do I have to memorize all these equations?
    No, but to understand them, which is basically the same.

Slides for the lectures

Chapter 01 - Introduction (pdf)
Chapter 02 - Linear learning machines (pdf)
Chapter 03 - Learning theory (pdf)
Chapter 04 - Neural networks (pdf)
Chapter 05 - Support-vector machines (pdf)
Chapter 06 - Deep learning (pdf)
Chapter 07 - Recurrent neural networks (pdf)
Chapter 08 - Reinforcement Learning (pdf)
Chapter 09 - Deep Reinforcement Learning (pdf)
Bonus - Introduction to the game of Go (pdf)

Exercises

To use Jupyter notebooks in the B202, follow these guidelines (pdf).

Exercise 01 - Introduction to Python and NumPy. (questions , solution )
Exercise 02 - Linear classification. (questions , solution )
Exercise 03 - Cross-validation. (questions , solution )
Exercise 04 - Multi-layer perceptron. (questions , solution )
Exercise 05 - Multi-layer perceptron on the MNIST dataset. (questions , solution )
Exercise 06 - Support-vector machines. (questions , solution )
Exercise 07 - Convolutional neural networks. (questions , solution )
Exercise 08 - Transfer learning. (questions , solution )
Exercise 09 - Reinforcement learning (skipped). (questions , solution )
Exercise 10 - Q-learning and Gridworld. (questions , solution )

Presseartikel

  • Neurologisch vernetzt

    Das Smart Start-Programm ermöglicht Masterstudenten ab September einen Einblick in Labore in ganz Deutschland – Auch TU-Student Alex Schwarz profitiert davon …

  • Roboter als soziale Wesen – wie passt das zusammen?

    Andreas Bischof von der Professur Medienkommunikation verbrachte im Rahmen seines Dissertationsprojektes beim Graduiertenkolleg CrossWorlds einen dreimonatigen Forschungsaufenthalt in den USA …

  • Wissbegierige Schüler weiter auf Erfolgskurs

    Zwei von der TU betreute Gymnasiasten errangen am 29. März 2014 Fachgebietssiege beim sächsischen Landeswettbewerb "Jugend forscht" und lösten damit ihre Fahrkarten zum Bundesfinale …

  • Raus aus den Federn, rein in die Hausarbeiten

    TU Chemnitz und Studentenwerk Chemnitz-Zwickau laden am 27. Februar zur „Langen Nacht der aufgeschobenen Hausarbeiten“ in die Mensa auf dem Campus Reichenhainer Straße ein – Start: 18:30 Uhr …