Advanced Artificial Intelligence

Jacky Baltes <jacky@cs.umanitoba.ca>

Important Updates

  1. Test 2 consists of three parts (worth 15 points each): Markov Decision Processes, Bayes Theorem and Translation, and Path Planning. Questions similar to Test 1. No definition or long answer questions. I want to know if you can understand and apply the various algorithms.
  2. Test 2 is on Wednesday, 6th April 2016 at 11:30 in EITC E2-304
  3. Class on Monday, 28th March 2016 - the TA will cover RRT slides. I will be back on Tuesday.
  4. Assignment 4 Path Planning is now available.
  5. Assignment 3 deadline extended until Tuesday, 29th March at 11:30.
  6. Assignment 3 Chinese Menu Translation is now available.
  7. Assignment 2 Deadline extended until Monday, 7th March 2016
  8. Added Myo Directional Keyboard option to the TensorFlow Assignment
  9. Comments about Midterm Test
  1. Topics: Everything discussed in class including 4 color theorem, constraint satisfaction, deep learning, overfitting, regression in TensorFlow.
  2. Most questions focus on algorithms and datastructures. I think of Input $I, Algorithm/Datastructure $A, and Output $O as 3 parameters of the system. I often give two and ask you for the third. Simplest case is: What is the output of algorithm $A/trace the execution of algorithm $A, given input $I? But I can also ask: Here is the output $O of algorithm $A, what was the input $I? Or here is $I and $O, what parameters were used for $A?  
  3. No long answer questions
  4. There are 4 - 6 questions on the midterm, grouped into three sections.
  5. 70% of the marks are on constraint satisfaction
  1. Class on Friday, 26th February, 2016 will be cancelled
  2. Test 1: Wednesday, 24th February, 2016 in EITC E2-304
  3. Announcement for Faculty of Science 2016 Pathways        to Exceptional Achievement
  4. The updated ROASS document is available
  5. Page opened Mon Jan  4 21:31:57 CST 2016

Content

This course covers a variety of topics in advanced artificial intelligence (AI). A collection of interesting problems and their modern approaches will be discussed. Students will implement AI systems to solve some of these problems.

Some possible sample problems are constraint satisfaction problems, reasoning about actions, natural language processing, simultaneous localization and mapping, and computer vision.

Lecture Slides

The following lecture slides are available.

  1. Constraint Satisfaction
  1. Four Colour Theorem
  2. Constraint Satisfaction
  1. Sudoku
  2. Cryptoarithmetic Puzzles
  1. Constraint Satisfaction Problems and Search
  2. Constraint Satisfaction Problems and Backtracking Search
  3. Constraint Satisfaction Problems and Forward Checking Search
  4. Example California Beach Volleyball
  5. Constraint Propagation
  1. Constraint Propagation and Semi-Magical Squares
  1. Constraint Satisfaction Problems and Problem Structure
  2. Heuristics for Constraint Satisfaction Problems
  3. Scheduling
  4. Performance of BT, Fwd Checking, and Const. Prop. on Nonograms
  1. Deep Learning
  1. Convolutional Neural Networks
  2. Regression
  3. TensorFlow
  4. Word2Vec
  1. Introduction to Bayesian Reasoning
  2. Markov Chains
  1. Markov Chains Sample Code
  2. Markov Chains Generator
  1. Motion and Path Planning
  1. Path Planner
  1. Rapidly Exploring Random Trees
  1. Rapidly Exploring Random Tree

Assignments

  1. Akari Puzzles and Constraint Satisfaction
  2. TensorFlow, Regression, Overfitting, and Myo
  3. English - Chinese Menu Translations
  4. Quadtree and RRTs

Previous Assignments

  1. Nonograms and Constraint Satisfaction
  1. Sample Solution

Old Exams