The course "Foundations of Artificial Intelligence VI" introduces introduces the mathematical foundations of supervised learning and discusses two learning algorithms in depth that are widely used in practice: decision tree learning and neural networks.  

Participants learn how to apply these algorithms to practical problems and how to evaluate the quality of a learned classifier. The course concludes with a discussion of the risks and challenges of machine learning.  

The course is based on the textbook by Stuart Russell and Peter Norvig: Introduction to Artificial Intelligence, 3rd edition, 2012 or 4th edition, 2020. The 3rd edition is also available in German: Stuart Russell und Peter Norvig: Künstliche Intelligenz - Ein Moderner Ansatz, 3. aktualisierte Auflage, Pearson 2012. 

The course is held in English language with German subtitles. 

Module Introduction to Machine Learning 

• The learning agent revisited 
• Types of learning 
• Ockham´s razor 

Module Decision Tree Learning 

• Boolean decision trees 
• Restaurant example 
• Overfitting 
• Theoretical properties 
• Recursive decision tree learning algorithm 
• Information Gain and Entropy 
• Entropy of a Boolean variable 
• Information Gain 
• Use in Decision Tree Learning 

Module Neural Networks 

• Introduction 
• History of neural networks 
• McCulloch-Pitts “Unit” Model of a Neuron 
• Activation functions 
• Feed-forward networks 
• Single layer feed-forward networks 
• Single layer feed-forward networks 
• Theoretical properties and limitations 
• Backpropagation algorithm 
• Take right-turn example 

Module Evaluation, Challenges and Risks 

• Machine learning workflow 
• Training and test data 
• Learning curve 
• K-fold cross validation 
• Evaluating a Boolean classifier 
• False Positives and false negatives 
• Accuracy, recall, specificity, precision 
• F1 Score 
• Risks and challenges of machine learning 
• Statistical vs. causal models 
• Wrong predictions with high confidence 
• Universal Adversarial Perturbations 

By the end of the course, you‘ll be able to 

• understand the theoretical foundations of machine learning algorithms, 
• evaluate a learned classifier 
• apply decision tree learning to practical problems 
• apply neural network learning to practical problems 
• understand challenges and risks of machine learning algorithms 

None.