Artificial Intelligence and Machine Learning - 2025 Embedded Robotics Department of Cybernetics and Robotics Faculty of Electronics, Fotonics, and Microsystems

Lecture topics and notes

These notes are under copyright. They can be used only for anybody's private purposes, and cannot be distributed or published, for example by copying and making available from other Web pages, or in any other way.

no	topic	slides
1 2	Searching in the state space Searching in games Searching in constraint satisfaction problems
3	Logic based methods
4	Logic programming in Prolog
5	Probabilistic representation: Bayesian networks
6	Making simple decisions: utility functions
7	Making complex decisions: Markov Decision Problems
8	Reinforcement learning algorithms
9 10 11	Machine learning: basic concepts and classification algorithms
12 13 14	Unsupervised learning
15	Computational theory of learning

Literature

1. S.J.Russell, P.Norvig, Artificial Intelligence A Modern Approach (Third Edition), Prentice-Hall, 2010, WWW
2. T.Mitchell, Machine Learning, McGraw Hill, 1997, WWW

1. Artificial intelligence courses with similar programs:
   • University of Waterloo, Canada, Introduction to Artificial Intelligence, Fall 2012, Pascal Poupart
   • M.I.T., Electrical Engineering and Computer Science, 6.034 Artificial Intelligence, Spring 2005, Tomás Lozano-Pérez, Leslie Kaelbling
   • M.I.T., Electrical Engineering and Computer Science, 6.825 Techniques in Artificial Intelligence, Fall 2002, Tomás Lozano-Pérez, Leslie Kaelbling

2. Polish language courses:
   • UJ, UW, PW, PP, kurs internetowy Sztuczna Inteligencja, Jaros³aw Arabas, Pawe³ Cichosz, Politechnika Warszawska
   • Uniwersytet Warszawski, Wydzia³ Matematyki, Informatyki i Mechaniki, Sztuczna inteligencja i systemy doradcze, Przedmioty fakultatywne dla II - V roku informatyki, semestr letni 2007/08, Arkadiusz Wojna

3. Search methods:
   • Slides for the search methods by prof.Tomás Lozano-Pérez and prof.Leslie Kaelbling from the M.I.T.:
     • Blind graph searching methods. Local copy:
     • Informed graph searching methods. Local copy:
     • A* algorithm and its properties. Local copy:
     • Methods for constraint problem solving. Local copy:
     • Methods for games. Local copy:

   • Slides for the search methods by prof.Andrew Moore from Carnegie-Mellon University. The course web page:
     • Blind algorithms. Local copy:
     • Hill-climbing algorithms. Local copy:
     • The A* algorithm. Local copy:
     • Algorithms for constraint problem solving. Local copy:

   • Slides for the search methods to accompany the book: Computational Intelligence: A Logical Approach, Poole, Mackworth and Goebel, Oxford University Press, 1998. The book web page:
     • Part 1. Local copy:
     • Part 2. Local copy:
     • Part 3. Local copy:
     • Part 4. Local copy:
     • Part 5. Local copy:
     • Part 6. Local copy:

   • search procedures library in C++

   • simulated annealing procedures library
     On-line demo: Local copy:

4. Textbooks and tutorials on Prolog:
   • A very simple and basic introductory Prolog tutorial.
   • A very good Prolog tutorial part 1, and part 2
   • Learn Prolog Now! another very good, comprehensive introduction to Prolog, An old version in a local PDF:
   • A short introductory textbook on Prolog programming.
   • Materials for a Prolog and logic programming course, slightly more advanced, with exercises for self-study, and solutions.

5. Systems for creating probabilistic belief networks:
   • JavaBayes written in Java, available in source, with examples and documentation, Java applet
   • Bayesian Network Toolbox (BNT), available in source, very comprehensive, covers many probabilistic models, does not have a GUI, does not work under Octave
   • MSBNx system developed at Microsoft Research runs on Windows (98, 2000, XP, and supposedly Vista)

6. Internet repositories of statistical data:
   UCI Knowledge Discovery in Databases Archive
   UCI Machine Learning Repository
   CMU StatLib Datasets Archive
   Kaggle database
   

7. Markov decision processes and reinforcement learning:
   • Sutton, Barto: Reinforcement Learning An Introduction, fundamental textbook on Markov processes and reinforcement learning, on-line version, second edition 2018

8. Supervised and unsupervised machine learning:
   • "A Few Useful Things to Know about Machine Learning", an overview paper on Machine Learning by Pedro Domingos, published in 2012 in the Communications of the ACM, PDF available: . Local copy:
   • "Top 10 algorithms in data mining", another influential paper on Machine Learning algorithms, published in 2007 in Knowledge and Information Systems, PDF available from many locations: . Local copy:

9. Neural networks and deep learning:
   • Michael A. Nielsen: Neural Networks and Deep Learning, Determination Press, 2015
   • Ian Goodfellow, Yoshua Bengio, Aaron Courville: Deep Learning, MIT Press, 2016

Final exam

The time and venue of the written examination will be announced during the first part of the semester.

“Grandson” tests

1. Tests are written, 3-minutes long, and can occur at any time during the lecture.
2. The scope of the test will be the current lecture, with a possible overlap of the preceding lecture.
3. This is a closed-book test. No books, notes, or computers can be used.
4. Calculators are permitted. During some tests, simple numerical calculations will be necessary. A cell phone, a calculator watch, or a PDA, will be allowed. But hand calculation is also possible.
5. Tests are graded from 1 to 3 points, with 1 point given for a blank page turned in.
6. No-show yields 0 points. No re-taking and no excuses.
7. The final “grandson” test score will be computed by dropping the single highest score and the single lowest earned score. 0 points result for absence is never dropped.

The Android app

If you do not have an Android smart phone, or have doubts about installing apps from an unknown source, don't worry.
You can learn how to quickly and reliably compute the “grandson” group manually, without the app.

Final exam waiver

The grades for the grandson test results will be interpreted as follows:

“Grandson” test results

The Project class

The schedule of project assignments is given below. Each assignment will be explained and discussed in the project class at the time of its start.

The assignments must be worked out individually. It is not allowed to share solutions with colleagues, or submitting results which are not the author's own. It is allowed to use any published resources, as long as they are properly credited and referenced in the report.

Either a written PDF report, or the code worked out, or both, must be turned in through the Moodle system. The detailed submission requirements for each assignment are given in the Moodle system, along with the grading criteria. However, the details of the projects are only provided here.

Grading the project class

The Project assignment schedule

	subject	descr.	tools	start	due	deadline
1.	Heuristic search - checkers		Checkers program + Java	March 12	April 1	April 8
2.	Logical reasoning - wumpus		Jovolog simulator + Prolog	March 26	April 15	April 22
3.	Bayesian networks modeling and decision making		many available, see description	April 9	April 29	May 6
4.	Markov decision problems and reinforcement learning		programming language to choose: Python3, Java, C/C++ Russel+Norvig 4x3 world data file assignment 4x4 world data file	April 23	May 13	May 20
5.	Classifier machine learning		Python3 scikit, many other libs data: text database	May 7	May 27	June 3
6.	Unsupervised learning		Python scikit, many other libs data: image database	May 21	June 10	June 17