Deep Learning - methods and applications, 7.5 Credits

Contents

The course introduces deep learning techniques and theories with a focus on practical applications. In the course a number of deep learning models will be covered, such as convolutional neural networks, recurrent neural networks and reinforcement learning. Possible uses and limitations of these algorithms will be discussed, and their implementation will be investigated using programming tasks.

Expected learning outcomes

After completing the course, the student should be able to:

Knowledge and understanding
*    describe central concepts and methods used in machine learning (FSR 1),
*    explain how different machine learning models generalize based on training examples (FSR 2),
*    categorize selected machine learning algorithms based on different aspects, such as supervised / unsupervised / semi-supervised /
     reinforcement learning, transfer learning, classification / regression / clustering (FSR 3),
*    explain strengths and limitations of selected algorithms / models for machine learning and how they can be applied in different applications
     such as image recognition and text classification (FSR 4).

Skills and abilities
*    design and implement appropriate machine learning solutions for given tasks (e.g. classifying images, analyzing text, or controlling agents in
     virtual environments) in a programming language suitable for machine learning (FSR 5),
*    apply modern development environments and software libraries to realize a machine learning task (FSR 6),
*    evaluate the performance of machine learning algorithms in terms of appropriate measurement values (e.g. accuracy, error rate, sensitivity,
     precision, recall, etc.) and compare them with theoretical expectations (FSR 7).

Judgement and approach
*    discuss advantages and disadvantages of different machine learning models based on a given application (FSR 8),
*    reason about the effects that can arise when applying machine learning-based systems on society, the environment and the labour market
     (FSR 9).

Required Knowledge

For admission to the course at least 7,5 hp in the area of Programming are required, preferable Python.

Form of instruction

The course is organized as a distance learning course. Teaching is conducted in the form of practical laboratory work with access to supervision. Pre-recorded lectures and/or seminar forums may be used.

Examination modes

Examination takes place through written exam, presentation of several programming assignments and a (seminar) discussion.
Written exam: Grading scale Fail (U), Pass(3), Pass with credit (4), Pass with distinction (5) (FSR 1, FSR 2, FSR 3)
Programming assignments: Grading scale Fail (U), Pass (G) or Pass with Distinction (VG) (FSR 2, FSR 4, FSR 5, FSR 6, FSR 7, FSR 8)
Discussion: Grading scale Fail (U) or Pass (G). (FSR 9)

The entire course is given one of the grades Fail (U), Pass(3), Pass with credit (4), Pass with distinction (5). To pass the course, all exams and compulsory parts must be passed.

To get a higher grade than 3 on the entire course:
- To get a grade of 5 on the course, grade VG from the laboratory work and grade 5 from the written exam are required.
- To get grade four on the course, either grade VG from the laboratory work and grade 4 from the written exam or grade G from the laboratory work and grade 5 from the written exam are required.

Students who have passed an examination may not undergo a re-examination for a higher grade.

For students who have not obtained the grade Pass, other examination sessions will be arranged. A student who for two consecutive examinations for the same course or sub-course has not been passed, has the right to have another examiner appointed, if there are no special reasons against this (Higher Education Ordinance chapter 6, 22 §). The written request for a new examiner shall be made to the Head of Department at Applied Physics and Electronics.

Deviations from the course syllabus examination form can be made for a student who has a decision on pedagogical support due to disability. Individual adaptation of the examination form should be considered based on the student's needs. The examination form is adapted within the framework of the expected syllabus of the course syllabus. At the request of the student, the course responsible teacher, in consultation with the examiner, must promptly decide on the adapted examination form. The decision must then be communicated to the student.

Students have the right to request that previous studies, or equivalent knowledge and skills acquired in a professional, work-related capacity, be validated and transferred into credits on an equivalent course or programme at Umeå University. Applications for credit transfer should be addressed to Student Services/Degree Evaluation Office. More information can be found at the Umeå University student web site (www.student.umu.se/english) and in Chapter 6 of the Higher Education Ordinance. Appeals may be made to the Higher Education Appeals Board (ÖNH) against a decision by the university not to approve an application for credit transfer (Higher Education Ordinance, Chapter 12), even in cases where only a part of the application has been rejected.

Other regulations

In the event that the syllabus ceases to apply or undergoes major changes, students are guaranteed at least three examinations (including the regular examination opportunity) according to the regulations in the syllabus that the student was originally registered on for a period of a maximum of two years from the time that the previous syllabus ceased to apply or that the course ended.