TH Köln

Master Digital Sciences

Documents for Study Program Accreditation

Module »Advanced Machine Learning« (AML)

Organizational Details

Responsible for the module
Prof. Dr. Gernot Heisenberg (Faculty F03)
Lecturer(s)
Prof. Dr. Gernot Heisenberg (Faculty F03), Prof. Dr. Konrad Förstner (Faculty F03)
Language
English
Offered in
Winter Semester (Duration 1 Semester)
Location
Campus Köln Süd, or remote
Number of participants
minimum 5, maximum 20
Precondition
none
Recommendation
Coding Skills in Python
ECTS
6
Effort
Total effort 180h
Total contact time
48h (24h lecture / 24h exercise)
Time for self-learning
132h
Exam
Project (during semester) in conjunction with expert talk
Competences taught by the module
Analyze Domains, Model Systems, Implement Concepts, Optimize Systems
General criteria covered by the module
Interdisciplinarity, Digitization, Transfer

Mapping to Focus Areas

Below, you find the module's mapping to the study program's focus areas. This is done as a contribution to all relevant focus areas (in ECTS, and content-wise). This is also relevant for setting the module in relation to other modules, and tells to what extent the module might be part of other study programs.

Focus Area ECTS (prop.) Module Contribution to Focus Area
Generating and Accessing Knowledge 3

This specialization recaps quickly the machine learning and especially deep learning principles, then dives deeper into current topics of the field.

Architecting and Coding Software 2

This module includes software development (notebooks)

Acting Responsibly 1

This module teaches how to deal with data and knowledge generating methods responsibly, acounting for ethics, privacy and security.

Learning Outcome

This specialization recaps quickly the machine learning and especially deep learning principles.

The student dives into the following topics

  • Advanced Feature Engineering Methods
    • Anomaly detection
    • Autoencoders
  • Generative Models
    • Variational Autoencoders
    • Generative Adversarial Networks
  • Explainable Machine Learning
  • Reinforcement learning

by filling their knowledge gaps between theory and practice while applying the methods in python solving natural language understanding and special computer vision real-world problems

for being able to apply modern machine learning methods in enterprises and research and understand the caveats of real-world data and settings.

Module Content

  • ML and DL principles (recap)
  • Advanced Feature Engineering Methods
    • Anomaly detection
      • Standardization,Box Plots,Correlation,DB-Scan Clustering,Isolation Forest,Robust Random Cut Forest
    • Autoencoders
      • feature selection and feature extraction
      • Latent variables and spaces
      • Image denoising
      • Missing value imputation / image impainting
      • Domain adaptation
  • Generative Models
    • Variational Autoencoders
    • Generative Adversarial Networks
  • Explainable Machine Learning
    • XAI methods and definitions
    • Partial Dependence Plots
    • Individual Conditional Expectation
    • Centered Individual Conditional Expectation
    • Derivative Individual Conditional Expectation
    • Shapley Values
    • Local Interpretable Model-agnostic Explanations (LIME)
  • Reinforcement learning
    • Definitions
    • Reinforcement control loop
    • Markov Decision process
    • Transition Probabilities
    • Discounted and Expected Return
    • Policies And Value Functions
    • The exploration-exploitation dilemma
    • Q-Learning
    • Deep Reinforcement Learning

Forms of Teaching and Learning

  • Lecture
  • Exercises and software development (notebooks)
  • Accompanying project work by analyzing data sets

Learning Material Provided by Lecturer

  • List of selected literature and web resources
  • Lecture slides
  • Video tutorials
  • Exercises and code tutorials
  • Example code and notebooks on github and Colab
  • Data sets and models

Literature

  • Ian Goodfellow, Yoshua Bengio und Aaron Courville: Deep Learning (Adaptive Computation and Machine Learning), MIT Press, Cambridge (USA), 2016. ISBN 978-0262035613 .
  • Neural Networks and Deep Learning by Michael Nielsen, ONline Book, http://neuralnetworksanddeeplearning.com/
  • Friedman, Jerome, Trevor Hastie, and Robert Tibshirani. “The elements of statistical learning”. www.web.stanford.edu/~hastie/ElemStatLearn/ (2009)
  • Doshi-Velez, Finale, and Been Kim. “Towards a rigorous science of interpretable machine learning,” no. Ml: 1–13. http://arxiv.org/abs/1702.08608 (2017)