Artificial Intelligence II

[Dates] [Contents] [References] [Slides] [Assessment]

Dates

Contents

Module Objective

Students will explore the state of the art in Artificial Intelligence.

Module Content

Topics will be selected from the following and others: constraint-based systems; probabilistic reasoning; machine learning; case-based reasoning; diagnostic systems.

Learning Outcomes

On successful completion of this module, students should be able to:

• Demonstrate a broad knowledge of modern AI theory and applications
• Discuss typical application scenarios and outline limitations
• Identify which AI technique is appropriate for solving a specific problem
• Apply and adapt the AI techniques covered in the lecture to concrete problems

References

• S. Russell, P. Norvig: Artificial Intelligence. A Modern Approach. Third Edition. Pearson, 2010.
• E. Alpaydin (2010): Introduction to Machine Learning. Second Edition. MIT Press.

Slides

1. 2012-01-04, slides 1-26 (long version) [introduction; machine learning; McCulloch-Pitts nets; perceptrons]
2. 2012-01-06, slides 27-59 (long version) [perceptron learning; unsupervised learning: clustering; reinforcement learning]
3. 2012-01-11, slides 60-91 (long version) [reinforcement learning; Oja's algorithm; multi-layer networks: computing XOR, bipolar encoding, towards learning: sigmoidal activation function]
4. 2012-01-18, slides 92-115 (long version) [backpropagation: general algorithm; adaptation to layered networks; complexity of the problem]
5. 2012-01-20, slides 116-146 (long version) [backpropagation: modifications and enhancements; support vector machines: definition; maximal margin separator; support vectors; slack variables and soft error]
6. 2012-01-25, slides 147-170 (long version) [support vector machines: kernel trick; kernel functions; hidden Markov models: defintion; Markov chains; evaluation: forward algorithm]
7. 2012-01-27, slides 171-194 (long version) [hidden Markov models: decoding: Viterbi algorithm; learning: Baum-Welch algorithm; constraint satisfaction problems: definition; comparison constraint programming and randomised search heuristics; graph colouring as CSP]
8. 2012-02-01, slides 195-226 (long version) [constraint satisfaction problems: definition; comparison constraint programming and randomised search heuristics; graph colouring as CSP; constraint graph; constraint propagation: node consistency; arc consistency; path consistency; strong k-consistency; Alldiff; bounds propagation]
9. 2012-02-03, slides 227-251 (long version) [solving CSP: backtracking; variable ordering: minimum-remaining-values, degree heuristic; value ordering; inference: forward checking; back-jumping; constraint learning; local search; 3-SAT as CSP]
10. 2012-02-08, slides 252-281 (long version) [time-limited local search for 3-SAT; "fast" exponential-time algorithms for 3-SAT; structure of CSPs: strongly structured CSPs: independent sub-problem, tree-structured CSPs]
11. 2012-02-10, slides 282-301 (long version) [structure of CSPs: weakly structured CSPs: cut sets, tree decompositions; evolving local search heuristics for solving CSPs; rational agents and uncertainty]
12. 2012-02-15, slides 302-333 (long version) [inference; independence; Bayes's theorem; Bayesian networks; reading Bayesian networks; constructing Bayesian networks]
13. 2012-02-17, slides 334-354 (long version) [Bayesian networks: Markov blanket; canonical distributions: deterministic nodes, noisy-or; continuous variables: discretisation, canonical parametric distributions, canonical non-parametric representations; hybrid Bayesian networks]
14. 2012-02-22, slides 355-379 (long version) [exact inference in Bayesian networks: enumeration, variable elimination; approximate inference in Bayesian networks: direct sampling; rejection sampling; likelihood weighting]
15. 2012-02-24, slides 380-402 (long version) [approximate inference in Bayesian networks: Gibbs sampling; utility theory: rational agents; preference relations; utility axioms; expected utility; utility functions; preference elicitation; utility of money]
16. 2012-02-29, slides 403-420 (long version) [utility of money; certainty equivalence; insurance premium; post-decision disappointment: optimiser's curse; human judgement: Allais paradox; Ellsberg paradox; source of irrational decisions; multiattribute utility functions]
17. 2012-03-02, slides 421-445 (long version) [multiattribute utility functions; stochastic dominance; decision networks; value of information; information-gathering agents; Markov decision processes; policies; additive rewards]
18. 2012-03-07, slides 446-464 (long version) [Markov decision processes: discounted rewards; utilities of policies and states; computing optimal policies: value iteration; policy loss]
19. 2012-03-09, slides 465-466 [in-class test]
20. 2012-03-14, slides 467-486 (long version) [solutions in-class test; optimal policies for Markov decision processes: policy iteration]
21. 2012-03-16, slides 487-506 (long version) [partially observable Markov decision processes: definition, belief state update, policies, value iteration]
22. 2012-03-21, slides 507-519 (long version) [summary of CS4619]
23. 2012-03-23, slides 520-538 (long version) [summary of CS4618]

Continuous Assessment

• weekly assignments handed out each Friday, due the following Wednesday, discussed the next Wednesday
  • Assignment 1, no written solutions due, solutions discussed 11.01.2012
  • Assignment 2, written solutions due 11.01.2012, solutions discussed 18.01.2012
  • Assignment 3, written solutions due 18.01.2012, solutions discussed 25.01.2012
  • Assignment 4, written solutions due 25.01.2012, solutions discussed 01.02.2012
  • Assignment 5, written solutions due 01.02.2012, solutions discussed 08.02.2012
  • Assignment 6, written solutions due 08.02.2012, solutions discussed 15.02.2012
  • Assignment 7, written solutions due 15.02.2012, solutions discussed 22.02.2012
  • Assignment 8, written solutions due 22.02.2012, solutions discussed 29.02.2012
  • Assignment 9, written solutions due 29.02.2012, solutions discussed 07.03.2012
  • Assignment 10, written solutions due 07.03.2012, solutions discussed 14.03.2012
  • Assignment 11, written solutions due 14.03.2012, solutions discussed 21.03.2012
• In-Class Test 09.03.2012

last change: 22.03.2012