[home]

Introduction to AI (2023/2024)

Tutorial for Introduction to Artificial Intelligence

This page is intended for both of the English tutorials.

Textbooks

The lecture is based on Russell and Norvig, Artificial Intelligence: A Modern Approach, 4th Edition (Prentice Hall, 2020)

Organization

To earn the credit, it is necessary to earn enough points for solving programming oriented assignments. Each assignment is worth at least 10 points. The threshold for passing is $7 \times \text{number of assignments}$. The expected number of assignments is $10$, but this may change due to some unforeseen circumstances.

All of the assignments will be posted on ReCodEx. The source codes to solve the assignments are hosted on Git. All of the assignments are implemented in Python and due to the automatic checks performed by ReCodEx it is necessary to adhere to the provided templates. The deadline will always be around 2 weeks, but for the exact dates, see ReCodEx.

Topics of assignments:

• Heuristics for A*
• Total coloring using CSP
• Total coloring using SAT
• Transporting boxes using PDDL
• Minesweeper using probability
• Localization using MDP
• Navigation using Bellman equations
• Simple game using Minimax
• Predicting diabetes using decision trees (report only)
• Article classification using NN (report only)

It is NOT mandatory to attend the classes, however, it is highly recommended as the topics discussed during the class may help you with solving the assignments.

It is forbidden to share the code of your solutions with your colleagues. On the other hand, it is allowed to discuss the approach to solving the assignments.

If someone feels that they are missing the opportunity to earn the credit, do not be afraid to contact me. Such cases will be dealt with individually based on your approach to the exercise.

Lectures

Here will be posted a brief summary of what happened in each lesson. If applicable, a PDF with exercises calculated during the lesson will also be posted.

20.2., 21.2.

• Introduction, resources, connection to recodex.
• Discussion about AI. What to expect from the class and what NOT to expect.
• Kaggle competitions, AIcrowd challenges.

27.2., 28.2.

• Search algorithms, search space vs. the input graph.
• Graph search vs. tree search.
• A*, admissible and monotonous heuristics.

4.3., 6.3.

• CSP
• Solving - backtracking, forward check, look ahead, AC-3
• Modeling - Sudoku, graph coloring, SEND+MORE=MONEY.

12.3., 13.3.

• SAT
• Why we want to use CNF, why not to use DNF.
• Solving - DPLL, unit propagation, pure literal elimination
• Modeling - at least one, at most one, sudoku, graph coloring

19.3., 20.3.

• Planning
• Modeling in PDDL. Transport car with 2 boxes.
• Tower of Hanoi, N-puzzle, Hamiltonian cycle.

26.3., 27.3.

• Probability reasoning
• Bayesian rule, conditional probability, chain rule.
• Wumpus world and connection to minesweeper.
• Continuing with Bayesian networks next time. The deadline for homework is 3 weeks this time.

2.4., 3.4.

• Bayesian networks
• Enumeration and variable elimination.

9.4., 10.4.

• Markov chain (MC) and hidden Markov model (HMM).
• Prediction of sunny/rainy days by MC.
• Filtering and smoothing for HMM. Nice slides with solved examples may be found here.

16.4., 17.4.

• MDP.
• Value iteration, policy iteration.

23.4., 24.4.

• Games.
• Turn-based games, minimax, alpha-beta. What are disadvantages of minimax and how to solve them.
• Single turn games, game theory, dominant strategy, Nash equilibrium, pure and mixed strategies.

30.4., 15.5.

• First part of machine learning.
• Regression, k-means, decision trees (will finish next time).

7.5., 22.5.

• Second part of machine learning.
• Neural networks, reinforcement learning.

21.5.

• Bonus lecture.
• Modern techniques. MCTS, Alpha[Go, ZeroGo, Zero], transformers.