DM568: Deep Learning

Internal Course Code

DM568

Comment

The course is part of 'Summer School' and is also for foreign students, held in week 32-33.
The course is co-read with DS833: Deep Learning for Data Science (5 ECTS) summer course.

Limited number of students

The course has a limited number of students enrolled. The following criteria, in the order listed below, are emphasized when allocating spots:

Students not enrolled as SDU
Students enrolled in IMADA programs
Other students

If there are ties in ranking among students within groups 1 to 3, selection will be based on the time of registration (first-come, first-served).

The academic departments at the Faculty of Science handle the prioritization process, and a waiting list will be established. Students who are not admitted to the course but are placed on the waiting list will be notified by the faculty. The waiting list will not carry over to the following year.

It is important to attend the first day of the course or notify the instructor, as the course has a waiting list.

Derogation from the general rules for cancellation

Cancellation of registration for this course is not permitted from the start of the semester and 21 days onwards, where cancellation is generally allowed. Therefore, registration for the course will be binding. Reference is made to the Collection of Rules for University of Southern Denmark regarding registration for course elements and exams §5, section 3-5.

Entry requirements

The course cannot be taken by students who are enrolled in the Data Science master program.
The course cannot be followed by students who have passed either DM873: Deep Learning (10 ECTS), DS809: Deep Learning (5 ECTS), AI506 or DS833.

Academic preconditions

Students taking the course are expected to:

Have competences in programming in Python
Basic understanding of linear algebra
Have knowledge corresponding to two years of Bachelor's studies in computer science

Participant limit

Course introduction

Machine learning has become a part in our everyday life, from simple product recommendations to personal electronic assistants to self-driving cars. More recently, through the advent of potent hardware and cheap computational power, “Deep Learning” has become a popular and powerful tool for learning from complex, large-scale data.
In this intensive course, we will discuss the fundamentals of deep learning and its application to various different fields. We will learn about the power but also the limitations of these deep neural networks. Finally, we will also dive into the technical and mathematical background of those methods, how they work and why they work in certain scenarios.
At the end of the course, the students will have significant familiarity with the subject and will be able to apply the learned techniques to a broad range of different fields.

Expected learning outcome

The learning objectives of the course is that the student demonstrates the ability to:

Describe the principles of deep neural networks in a scientific and precise language and notation
Analyze the various types of neural networks, the different layers and their interplay
Discuss the feasibility of deep learning approaches to concrete problems
Describe the theoretical mathematical foundations of the field
Implement and apply deep learning frameworks for solving concrete problems
Utilize state-of-the-art deep learning frameworks for implementing deep neural networks

Content

The following main topics are contained in the course:

feedforward neural networks
recurrent neural networks
convolutional neural networks
backpropagation algorithm
mathematical background of the methods
regularization
advanced neural network optimization techniques

Literature

See itslearning for literature references.

Examination regulations

Exam element a)

Timing

July/August

Tests

Project with oral defence

EKA

N330026102

Assessment

Second examiner: None

Grading

Pass/Fail

Identification

Full name and SDU username

Language

Normally, the same as teaching language

Examination aids

All common aids allowed

ECTS value

Additional information

The oral defence takes place in the end of the teaching period.

Reexam: still project with oral defence. The project is a new assignment to be submitted, usually in September month, with subsequence oral defence. The oral defence can be done virtually depending on the home location of the student.

Indicative number of lessons

60 hours per semester

Teaching Method

Planned lessons:

Total number of planned lessons: 60

Hereof:

Common lessons in classroom/auditorium: 42

Common lessons in laboratory: 18

The course will consist of frontal lectures supported by a discussion session followed by an accompanying exercise or lab session. Here, the students are supposed to directly transfer their acquired knowledge into practice by solving problems of increasing difficulty.

Other planned teaching activities:

The students are supposed to work together in small groups and review the course material and read accompanying chapters for the next day of lectures.

Teacher responsible

Name	E-mail	Department
Richard Röttger	roettger@imada.sdu.dk	Data Science

Timetable

Odense

Show full time table

Administrative Unit

Institut for Matematik og Datalogi (datalogi)

Team at Registration

NAT

Offered in

Odense

Recommended course of study

Profile	Education	Semester	Offer period

Transition rules

Transitional arrangements describe how a course replaces another course when changes are made to the course of study.

If a transitional arrangement has been made for a course, it will be stated in the list.

See transitional arrangements for all courses at the Faculty of Science.