Artificial Intelligence II HWS 12/13

 

Important:

  • Lectures will be held in B 6 23-25, Bauteil A, room A 3.02
  • Sign up for the ILIAS group! (this is the only source you will need during the semester)

 

Course Description

The class will provide and advanced overview of the state-of-the-art principles and methods of artificial intelligence. The covered areas will be logics for artificial intelligence, uncertain reasoning, statistical relational learning, and optimization algorithms. Students will learn how to apply the learned methods to solve meaningful problems in industry and economics. For instance, constructing a telecommunication network at low cost while guaranteeing stability if particular connections get damaged, and efficient traffic management are some of the applications that will be addressed in the class. For students of economics we  might provide additional lectures to catch up with the education of computer scientists (e.g. logic and programming usually need to be introduced to economists).

  • Probabilistic Reasoning & Statistical Relational Learning: Bayesian and Markov Networks, Markov Logic
  • Decision Making & Planning: Markov decision processes, Reinforcement Learning
  • Optimization Problems & Methods used for probabilistic reasoning
  • Real world applications: (Semantic) Web applications, Applications in economics and industry

Course level: Master and Diploma; Course language: English;
Prerequisites: Basic programming skills; Basic knowledge of formal logic; Basic knowledge of probability theory

 

Organization

  • Lectures: Thursdays 15:30-17:00 Location: B 6, 23-25 Bauteil A - Room  A 3.02

  • Exercise: Wednesdays 13:45-15:15 Location:  B 6, 23-25  Bauteil A - Room  A 3.02 

The first day of class will be Thurday Sept. 6th  

 

Homework, Projects and Final Exam

There will be homework assignments to prepare you for the exam but the main part of the course will be a semester project and a presentation together worth 60% of the final grade. This is expected to be a project going deeper into an area that is related to the topics covered in class. Projects may be done individually or in teams of two students. There are essentially two possible choices:

  1. Review projects: Students are supposed to review existing literature and write a comprehensive 30 page paper about the state-of-the-art in the area. 
    Example: „Learning the structure of Bayesian networks from databases.“
  2. Programming projects: Students are supposed to develop algorithms that employ methods related to the topics covered in class to solve a specific class of problems. The project should involve the creation of software (code and documentation) and a short paper (about 5 pages) briefly explaining the approach. Programs should include well-documented code and the output should demonstrate its processing. 
    Example: „Using a Prolog-System for Reasoning in light-weight Ontologies“

Papers should be formatted according to the IEEE conference template.

There will be a final exam worth 40% of the final grade. It will cover the material taught in class lectures.

 

Presentation and Commentators

The class will be divided into groups to work on the research projects. There will be in-class presentations worth 20% of the final grade. At the end of the course, each project group will briefly present their results in about 20 minutes. Before the presentation, the entire class will read the related papers.

For each presentation, one or more students will be assigned as commentators. The commentator will make a brief statement after the main presentation, sharing their own ideas and responding to the issues raised during the presentation.  

 

Grading

Composition of the Grade:

  • In-class presentations and remarks/questions of commentators will count 20%,
  • the final project will count 40%, and
  • the final exam also 40%.

Of course, different group members may make different types of contributions but the overall contributions are expected to be the same for each team member to receive the same project score. If we determines that there were significantly different levels of contribution, relative contributions may be used to weight final grades.  

 

Literature

Main Textbooks:

  • Stuart Russel and Peter Norvig: Artificial Intelligence - A Modern Approach, Prentice Hall 2003;http://aima.cs.berkeley.edu/ (2nd or 3rd edition)
  • Luc de Raedt: Logical and Relational Learning, Springer Verlag 2008
  • Markov Logic networks, Machine Learning, 2006 (link)

 

Syllabus

(06.09.2012) Introduction Chapter: 1 

  • Organisation
  • History of AI
  • Motivation
  • Examples

(13.9.2012) Logical Agents Chapter: 2 & 7 

  • Intelligent Agents (2.1-2.3)
  • Knowledge Based Agents (7.1)
  • The Wumpus World (7.2)
  • Logic  (7.3)

(20.9.2012) Propositional Logic Chapter: 7 

  • Propositional Logic – A Very Simple Logic (7.4)
  • Propositional Theorem Proving (7.5)
  • Effective Model Checking (7.6)
  • Agents based on Propositional Logic (7.7)

(27.9.2012) First Order Logic Chapter: 8 

  • Representation revisited (8.1)
  • Syntax and Semantics of First-Order Logic (8.2)
  • Using First-Order Logic (8.3)

(4.10.2012) Inference in First order Logic Chapter: 9

  • Propositional vs. First-Order Inference (9.1)
  • Unification and Lifting (9.2)
  • Forward Chaining (9.3)
  • Backward Chaining (9.4)
  • Resolution (9.5)

(11.10.2012) Uncertainty Chapter: 13

  • Acting under Uncertainty (13.1)
  • Basic Probability Notation (13.2)
  • Inference Using Full Joint Distributions (13.3)
  • Independence (13.4)
  • Bayes' Rule and its Use (13.5)
  • The Wumpus World Revisited (13.6)

(18.10.2012) Probabilistic Reasoning Chapter: 14

  • Representing Knowledge in an Uncertain Domain (14.1)
  • The Semantics of Bayesian Networks (14.2)
  • Efficient Representation of Conditional Distributions (14.3)Supplemental video lecture:

    1. videolectures.net/mlss05au_roweis_pgm/

(25.10.2012) Probabilistic Reasoning Chapter: 14 

  • Exact Inference in Bayesian Networks (14.4)
  • Approximate Inference in Bayesian Networks (14.5)

(01.11.2012) Statistical Relational Learning  (DeRaedt, MLpaper) 

(08.11.2012) Making Simple Decisions: Chapter 16 

(15.11.2012) Making Complex Decisions: Chapter  17

(22.11.2012) Applications

(29.11.2012) Presentations

  1. Some tips on giving an academic talk

Exercise Sheets and Slides

Homework assignments and recitation slides are available in ILIAS.

Forum

For questions, suggestions, and other comments please use  ILIAS. Please do not forget to check it regularly for comments concerning organizational issues around this lecture.