Probabilistic Models

Main themes and learning objectives »
582636
5
Algorithms and machine learning
Advanced studies
This course provides an introduction to probabilistic modeling from a computer scientist"s perspective. Many of the research issues in Artificial Intelligence, Computational Intelligence and Machine Learning/Data Mining can be viewed as topics in the "science of uncertainty," which addresses the problem of optimal processing of incomplete information, i.e., plausible inference, and this course shows how the probabilistic modeling framework forms a theoretically elegant and practically useful solution to this problem. The course focuses on the "degree-of-belief" interpretation of probability and illustrates the use of Bayes" Theorem as a general rule of belief-updating. As a concrete example of methodological tools based on this approach, we will study probabilistic graphical models focusing in particular on (discrete) Bayesian networks, and on their applications in different probabilistic modeling tasks.

Exam

25.02.2013 16.00 B123
Year Semester Date Period Language In charge
2013 spring 15.01-21.02. 3-3 English Petri Myllymäki

Lectures

Time Room Lecturer Date
Tue 16-18 D122 Petri Myllymäki 15.01.2013-21.02.2013
Thu 16-18 D122 Petri Myllymäki 15.01.2013-21.02.2013

Exercise groups

Group: 1
Time Room Instructor Date Observe
Wed 16-18 B119 Joonas Paalasmaa 21.01.2013—22.02.2013

Information for international students

The course will be held in English.

General

This course belongs to the Algorithms and Machine Learning sub-pogramme in the Master's programme of the department, and together with 582637 Project in Probabilistic Models (2 cr), it forms one of the three optional courses of the sub-programme.

For students in the old Intelligent Systems specialisation area: this course replaces, together with the project work 582637 Project in Probabilistic Models  (2 cr), the course Three Concepts: Probability (6 cr).

The format of the course in Spring 2012 follows roughly the format of the course in 2011.

The course is an introductory course, and only elementary knowledge on probability theory is required as a prerequisite. Different parts of the course, however, have different requirements with respect to the mathematical machinery needed to apply the concepts in question. Typically some analysis and elementary mathematical statistics is required. At the very least we assume that the participants are familiar with topics covered in the courses 582630 Design and analysis of algorithms (4 cr) and 582631 Introduction to machine learning (4 cr).

 

Completing the course

N.B. The plan below is still tentative and subject to changes!

There will be weekly exercises and a final exam. Solutions to the weekly exercises need to be delivered to the course assistant before each exercise session (by paper or by email).

The maximum number of points that can be earned from the exercises is 12 and from the course exam 48, so the total maximum is 60 points (and you need 30 points in total and 24 points from the exam to pass the course).  In the exam, the use of a basic calculator is allowed but no other electronic devices or written material. The exam dates are as follows:

  • Final exam (max 48 exam points, excercise points valid): 25.02. at 16:00
  • Renewal exam (max 48 exam points, excercise points valid):: 16.04. at 16:00
  • Separate exam (max 60 exam points, no excercise points): 14.06. at 16:00

Always double-check the times and locations of the exams from the department web pages.

The results are out. Thank you for everybody for participating - if you wish to discuss your grade, please contact the course instructor. Your comments on the course (contents, format, etc.) are  greatly appreciated, so please use the anonymous web form for sending us feedback  (especially verbal comments are welcome).

Lecture Schedule

Tue 15.01. at 16-18: Overview of the course, administrative issues, background and motivation. Lecture slides Part I: 1-24.

Thu 17.01. at 16-18: More background and motivation. Lecture slides Part I: 25-73. Part II: 1-25.

Tue 22.01. at 16-18: Bayesian inference. Lecture slides Part II: 26-65 (updated 22.01.2013)

Thu 24.01. at 16-18: Introduction to Bayesian networks. Lecture slides part III:1-33.

Tue 29.01. at 16-18: The Bayesian network model family. Lecture slides part III:34-52 (updated 29.01.2013).

Thu 31.01. at 16-18: Inference in Bayesian networks. Lecture slides part IV: 1-18 (updated 31.01.2013).

Tue 05.02. at 16-18: Inference in Bayesian networks. Lecture slides IV:19-55 (updated 05.02.2013).

Thu 07.02. at 16-18: Inference in Bayesian networks. Lecture slides IV:56-81 (updated 07.02.2013).

Tue 12.02. at 16-18: Learning in Bayesian networks. Lecture slides V: 1-23.

Thu 14.02. at 16-18: Learning Bayesian networks. Lecture slides V: 24-46 (updated 14.02.2013).

Tue 19.02. at 16-18:30: Miscellenious topics (missing data, latent variable models, undirected graphical models). Lecture slides V: 47-56. Lecture slides VI.

Thu 21.02. at 16:00-17:00:  Invited talk by Eric Brown: "Watson: The Jeopardy! Challenge and Beyond".

Thu 21.02. at 17:15-18:00: Q&A with Eric Brown.

 

Exercises

Course assistant: Joonas Paalasmaa

Grading: The maximum number of points from the exercises is 12. The solutions are graded on a scale of 0-2 "effort points":

  • 0: No significant progress (e.g., lack of proper effort, or a serious misunderstanding).
  • 1: A decent effort but perhaps some key element(s) missing.
  • 2: Correct (or almost correct) solution.

The effort points are scaled to exercise points in such a way that 0 effort points yields 0 exercise points and 80% of maximum effort points yields the maximum of 12 exercise points. Return your solution to Joonas before each Wednesday at 12 0'clock noon (by email to joonas.paalasmaa at helsinki.fi, cc: petri.myllymaki at cs.helsinki.fi), and only solutions submitted before this deadline are eligible to exercise points (read the detailed instructions on the exercise sheet). The exercises will not be graded until the end of the course.

Summary of effort points after sessions 1-3.

NB. In order to be allowed to send a modfied version of a solution after the exercise session, note that

  1. You can only modify a solution that you submitted before  the deadline before the exercise session (i.e., you can NOT send a modified solution to an exercise you did not work on before the exercise session)
  2. You can only send a modified version for one solution, and ONLY if you attended the exercise session.

ALSO: Please use the subject line of your email submission as the name of the submitted file as well.

Schedule of the exercise sessions:

 

Literature and material

The primary material is the lectures notes.

Additional material: