Tsinghua University · 2026
LOGIC · DATA · LEARNING
Logic, Data Examples and Learning
This five-lecture course explores the connections between logic, data examples, and learning. It begins with the role of examples in explaining and characterizing concepts, then reverses the direction to study algorithms that construct fitting concepts from labeled data. The course continues with generalization, PAC learning, and Occam algorithms, before turning to graph neural networks and interactive exact learning.
COURSE MATERIALS
Lectures & Slides
01
Setting the Stage
The Power and Limitations of Data Examples
02
Fitting Algorithms
Fitting Algorithms for Logical Concept Classes
03
Generalization
When Do Fitting Concepts Generalize to Unseen Examples?
04
Graph Neural Networks
Graph Neural Networks and Their Connections to Logic
05
Interactive Learning
Learning by Asking the Right Questions