lunduniversity.lu.se

Third-Cycle Courses

Faculty of Engineering | Lund University

Details for the Course Syllabus for Course EDA055F valid from Autumn 2019

Printable view

General
Aim
  • Graphical Models, Bayesian Learning and Statistical Relational Learning is a core course within the Wallenberg AI, Autonomous Systems and Software Program (WASP), more specifically within the the WASP-AI Graduate School track. The purpose of the course is to give a deepened understanding of the type of models and machine learning methods mentioned in the course title. These are different from, e.g., deep learning or reinforcement learning approaches to a degree, that motivates a respective specific course to broaden the participants' knowledge within the general core topic of machine learning within WASP-AI.
Contents
  • The course is structured in three modules, dealing with graphical models, algorithms for doing inference in and learning of them, and the combination of logical and probabilistic approaches to reasoning.

    Topics to be discussed include probabilistic graphical models, causal models, interventional distributions and structural learning algorithms (module 1); Markov chain Monte Carlo, approximate message-passing, and variational inference, with a particular emphasis on inference in probabilistic graphical models (module 2); Generalized syntax and semantics of propositional and (especially) predicate logic, as well as major results about algorithmic decidability and efficiency for logical formalisms (module 3).
Knowledge and Understanding
  • For a passing grade the doctoral student must
  • - know the terms, concepts and mathematical / logical expressions used in the presented material,
    - be able to understand and explain the presented concepts,
    - be able to use the presented concepts to express a given problem with them
    - know the conceptual and mathematical tools needed to solve given problems
Competences and Skills
  • For a passing grade the doctoral student must
  • - be able to solve given problems with the tools presented in the course, at least to a certain level of complexity
    - decide upon the applicability of a certain tool in a given context outside the scope of the course
    - transfer the learned concepts and tools to problems outside the scope of the course where applicable
Judgement and Approach
  • For a passing grade the doctoral student must
  • - be able to judge the advantages, scope of applicability and limitations of the presented tools and concepts in a given (research related) context
    - be able to evaluate the consequences of applying the tools, methods and concepts discussed in the course outside the course scope
Types of Instruction
  • Lectures
  • Laboratory exercises
  • Exercises
  • Self-study literature review
  • Miscellaneous
  • Each of the course modules is designed individually, containing some combination out of lectures, exercises, practical exercises, study material / reading advise, and homework assignments . A module consists of a two-day physical meeting with lectures and some form of exercises, as well as some homework assignments that the students can work on prior / after the meeting.
Examination Formats
  • Written assignments
  • Miscellaneous
  • The examination criteria aside the active participation in the physical meeting for each module are set by the teachers responsible for the respective module and communicated prior to or at the physical meeting for the module.
  • Failed, pass
Admission Requirements
Assumed Prior Knowledge
  • Specific assumed prior knowledge is communicated through the course website per module, which is available well in advance. Generally the course requires at least some knowledge in probabilistic representation and reasoning, graphical models, and first order logic. Students are expected to have a background in Computer Science, Mathematics, Engineering Physics, Electrical Engineering or a closely related topic.
Selection Criteria
Literature
  •  
  • Lecture material (slides, articles, and literature recommendations) and hand-in assignments are distributed via the course homepage or the used and announced course content management system respectively.
Further Information
Course code
  • EDA055F
Administrative Information
  • 2019-10-08
  • Professor Thomas Johansson

All Published Course Occasions for the Course Syllabus

1 course occasion.

Start Date End Date Published
2019‑09‑01 (approximate) 2020‑01‑15

Printable view