Machine learning, 7.5 Credits

Contents

The course introduces the basic grounding in concepts and a range of model based and algorithmic machine learning methods, as well as practical design and evaluation of some machine learning solutions.
 
The course is divided into three parts:
Part 1: Principles, 4 credits.
This part introduces the background and important applications of machine learning. The following topics will be covered: support vector machines, naive bayes, k-nearest neighbor, decision trees, neural networks, linear regression, clustering, hidden markov model, and reinforcement learning. Meanwhile, a series of important concepts and knowledge will be mentioned including bias/variance tradeoffs, generative/discriminative learning, kernel methods, parametric/non-parametric learning, graphic models, and deep learning. Some machine learning libraries (e.g. scikit- learn) and development tool will be briefly introduced.
 
Part 2: Seminar, 0.5 credits.
Research papers on some specific machine learning topics will be provided in advance to all the students for reading. In this part, the students are divided into groups and are required to give short presentations to explain and assess the assigned paper (i.e. the proposed solution, datasets, experimental design etc.).  One group representative can be chosen to present but all group members must be present and involved in discussion. The teacher and other students will ask questions and give feedback.
 
Part 3: Practice, 3 credits.
This part consists of 2-4 mandatory practical assignments. The students are required to apply some of the machine learning algorithms/models and frameworks/libraries from Part 1 into practice by solving the assigned problems in real applications (e.g. text classification, search engine, or recommendation system).
 

Expected learning outcomes

Knowledge and understanding
After completing the course the student will be able to:

• Explain and describe some concepts and methods central to machine learning such as classification, regression, clustering, bias/variance, kernel functions, and optimization. (FSR 1)
• Categorize selected machine learning algorithms from different aspects, such as supervised/unsupervised/semi-supervised, classification/regression/clustering, generative/discriminant, and parametric/non-parametric. (FSR 2)
• Explain the strengths and limitations of selected machine learning algorithms/models and how they can be applied in some different applications such as text classification, ranking, image recognition. (FSR 3)

Skills and abilities
After completing the course the student will be able to:

• Design and implement suitable machine learning solutions to given tasks (e.g. classify/cluster documents by topic) in state-of-the-art programming languages. (FSR 4)
• Apply some state-of-the-art development frameworks and software libraries in machine learning task realization. (FSR 5)
• Evaluate the performance of machine learning algorithms in terms of suitable measure metrics (e.g. accuracy, error rates, sensitivity, precision, recall etc.), and compare them with theoretical expectations. (FSR 6)

Values and attitudes
After completing the course the student will be able to:

• Discuss the impact on real applications and society of new technologies in machine learning. (FSR 7) 

Required Knowledge

Univ: To be admitted you must have (or equivalent) 90 ECTS-credits including 60 ECTS-credits in Computing Science or two years of completed studies within a study programme (120 ECTS-credits). In both cases, including at least 7.5 ECTS-credits in artificial intelligence (e.g. 5DV121) and at least 7.5 ECTS-credits in mathematical statistics, (e.g. 5MS045).

Proficiency in English equivalent to Swedish upper Secondary course English A/5. Where the language of instruction is Swedish, applicants must prove proficiency in Swedish to the level required for basic eligibility for higher studies.

Form of instruction

The course consists of lectures, seminars and practical assignments. In addition to scheduled activities also requires individual work with the material.
 

Examination modes

Assessment on the course consists of a series of assignments (50%), one seminar (15%) and one written exam (35%). We totally assign 1000 points to this course as follows:

Part 1: Principles - 350 points (FSR 1, 2, 3): The rating of Part 1 is any of the following: Fail (U) or Pass (G) and to get a pass grade the student must have at least 175P at the part. The assessment is done by a written exam in halls.

Part 2: Seminar - 150 points (FSR 2, 3, 7): The rating of Part 2 is any of the following: Fail (U) or Pass (G) and to get a pass grade the student must have at least 75P at the part.  The assessment is made during a mandatory seminar where the students are required to give short presentations to explain and assess the assigned paper (i.e. the proposed solution, datasets, experimental design etc.). The teacher and other students will ask questions and give feedback.

Part 3: Practice - 500 points (FSR 4, 5, 6): The rating of Part 1 is any of the following: Fail (U) or Pass (G) and to get a pass grade the student must have at least 250P at the part. The assessment consists of 2-4 mandatory assignments that are presented in written reports (e.g. code, solution description, and algorithm evaluation).
 
At the course as a whole one of the grades Pass with Distinction (5), Pass with Merit (4), Pass (3), or Fail (U) is given. To pass the course completely, all mandatory parts must be passed. The final score is determined from the collected score according to the following scale:
 
Pass with Distinction (5): p> = 900
Passed with Merit (4): 900> p> = 750
Pass (3): 750> p> = 500
Fail (U): 500> p

An additional seminar will be arranged for students who fail to attend the first seminar. Students who are absent from both seminars have to attend the seminar in the next year to get the corresponding credits to pass the course. 

 A student who has passed an examination may not be re-examined.

A student who has taken two tests for a course or segment of a course, without passing, has the right to have another examiner appointed, unless there exist special reasons (Higher Education Ordinance Chapter 6, section 22). Requests for new examiners are made to the head of the Department of Computing Science.

Examination based on this syllabus is guaranteed for two years after the first registration on the course. This applies even if the course is closed down and this syllabus ceased to be valid.

TRANSFER OF CREDITS
Students have the right to be tried on prior education or equivalent knowledge and skills acquired in the profession can be credited for the same education at Umeå University. Application for credit is submitted to the Student Services / Degree. For more information on credit transfer available at Umeå University's student web, www.student.umu.se, and the Higher Education Ordinance (Chapter 6). A refusal of crediting can be appealed (Higher Education chapter 12) to the University Appeals Board. This applies to the whole as part of the application for credit transfer is rejected.

Other regulations

This course may not be used towards a degree, in whole or in part, togehter with another course of similar content. If in doubt, consult the student counselors at the Department of Computing Science and / or the program director of your program.