Instructor: |
Alex Thomo |
Phone: |
(250) 721-8659 |
Office: |
EOW 315 |
Office Hours: |
Monday/Thursday 1:00-2:30 |
Email: |
thomo@cs.uvic.ca |
TA: |
Alan Ezust |
Email: |
sae@uvic.ca |
Course Outline: |
Link |
Text: S. Russell and P. Norvig,
Artificial Intelligence, a Modern Approach, 2nd Edition, Prentice
Hall, 2003
Languages: We will use Java and Prolog
for the programming part of assignments. Java for AI in other
universities: MIT, Stanford,
and McGill.
Midterm: Monday, Feb. 28, in class. Solutions.
Final Exam: Apr. 15, Solutions.
Assignments: There would be five
assignments, which will have both paper and programming questions.
The assignments would be team work wherein you are required to team up
with another classmate.
Source Code: Here is the book source
code. Also, you can download a jar file and put it in
the classpath. The UML diagrams (after reverse engineering) are: diagram1, diagram2, diagram3 (save
diagram3 and then zoom it).
The code for the Water Jugs problem is here. Also, check some
instructions
on how to implement a search problem.
WebBoard: Link
E-Submission: Link
Marks so far: Link
Assignment 1
Solutions
(missionaries-cannibals and water-jugs)
Assignment 2
Examples: Australia CSP
coloring (also download again aima.jar above). General framework
for games. Tic-tac-toe.
Solutions
(non-programming ones), Zebra,
Nim
Assignment 3
Theo
Theorem Prover (Linux) Theo Theorem Prover
(SunOS) Theo
Theorem Prover (Windows/DOS) Solutions(tar file)
Assignment 4
Solutions
Assignment 5
Lectures and Handouts:
Problem-solving.
- Overview of AI. History of AI. Application of AI. Topics to be
covered in the course. Slides(ppt)
- Solving Problems by Searching. Formulating problems. Searching for
Solutions. Measuring problem-solving performance. Uninformed Search
Strategies. Iterative deepening depth-first search. Comparing uninformed
search strategies. Avoiding Repeated States. Slides(ppt).
- Informed Search and Exploration. Informed (Heuristic) Search
Strategies. Greedy best-first search. A* search. Heuristic Functions.
Inventing admissible heuristic functions. Local Search Algorithms and
Optimization Problems. Hill-climbing search. Slides(ppt).
- Constraint Satisfaction Problems. Backtracking Search for CSPs.
Variable and value ordering. Propagating information through
constraints. Forward checking. Constraint propagation. Local Search for
Constraint Satisfaction Problems Slides(ppt).
- Games. Optimal Decisions in Games. Optimal strategies. The minimax
algorithm. Alpha-Beta Pruning. Imperfect, Real-Time Decisions.
Evaluation functions. Cutting off search. Slides(ppt).
Knowledge and Reasoning.
- An informal introduction to logic. Human Logic. Proofs. Unsound
Patterns. Soundness. Deduction. Formalization. Resolution Principle.
Application to Databases. Slides(ppt).
- Propositional Logic. Syntax. Semantics. Interpretations. Evaluation.
Reverse Evaluation. Validity, Satisfiability, Unsatisfiability.
Entailment. Semantic Reasoning Methods. Proofs. Resolution Principle. Slides(ppt).
- Conceptualization. First-Order Logic. Syntax. Semantics: Models.
Entailment. Herbrand's Models and Method. Slides(ppt).
- Converting to Clausal Form: Skolemization, Inseado. Substitutions
and Unification. Computing the Most General Unifier. Relational
Resolution. Examples. Slides(ppt).
- Examples, Strategies for resolution. Theo Theorem Prover. Slides(ppt).
- Horn Clauses. Ordered resolution. Directed Resolution: Forward and
Backward Resolution. Prolog. Slides(ppt).
Uncertain Knowledge and Reasoning.
- Handling Uncertainty. Belief and Probability. Atomic events. Joint
probability. Conditional probability. Inference by enumeration.
Independence. Bayes' Rule. Combining evidence. Conditional independence.
Slides(ppt).
- Bayesian networks. Compactness of Conditional Probability Tables
(CPT’s). Semantics. Constructing Bayesian networks. Noisy-OR. Exact
Inference in Bayesian Networks. Slides(ppt).
Machine Learning.
- Induction. Kinds of learning. Memory, averaging, and generalization.
Variety of Learning Methods. Data Mining. Complexity of hypothesis.
Learning DNF. Slides(ppt).
- Inferring rudimentary rules (1R). Dealing with numeric attributes.
The problem of overfitting. Statistical modeling. Naive Bayes
classifier. The “zero-frequency problem”. Missing values. Slides(ppt).
- Decision Trees. Entropy. Information gain. Highly-branching
attributes. Slides(ppt).
- Rules. Covering. Separate-and-Conquer (vs. Divide-and-Conquer).
PRISM algorithm. Slides(ppt).
- Mining Association Rules. Frequent Item sets. Generating item sets
efficiently. Slides(ppt).
- Credibility: Evaluating what’s been learned. Training and testing.
Cross-validation. Slides(ppt).
- Nearest Neighbor. Scaling. Voronoi partition of the space. KD-Trees.
K-Nearest Neighbors. Slides(ppt).
- More on Decision Trees. Numerical attributes. Slides(ppt).
- Regression: Locally weighted averaging. Kernel functions for
weighting. Regression Trees. Slides(ppt).
- Linear Separators. Hyperplane Geometry. Margin. Perceptron
algorithm. Slides(ppt).
- Neural Networks: Single Perceptron Unit. Beyond Linear Separability.
Multi-Layer Perceptron Learning. Soft Threshold. Sigmoid Unit. Gradient
Descent. Delta Rule. Backpropagation Algorithm. Training Neural Nets. Slides(ppt).
Review.
Notes: All assignments are due in class.
There will be a grade penalty for late assignments. All programs must have
adequate internal documentation.
|