DM873: Deep Learning

Comment

See Danish version

Entry requirements

The course cannot be chosen by students who have taken DM863: Deep Learning (5 ECTS), DS809: Deep Learning (5 ECTS), DM568: Deep Learning (summer school) (5 ECTS) or DS833..

Academic preconditions

Students taking the course are expected to:

To be profound programmers
Basic understanding of linear algebra

Course introduction

Machine learning has become a part in our everydays life, from simple product recommendations to personal electronic assistant 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 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. 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.

The course builds partly on the knowledge acquired in the course DM555 but can be taken by any Computer Science or Computational BioMedicine Master student.

In relation to the competence profile of the degree it is the explicit focus of the course to:

giving the competence to plan and execute a deep learning task by means of deep neural networks.
providing knowledge on the different types of deep learning approaches including their advantages and disadvantages.
transfer learned methods to new fields of applications.
challenges the student with real-life datasets and problem-solving skills

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
Describe the feasibility of deep learning approaches to concrete problems
Understand the theoretical mathematical foundations of the field
Apply deep learning frameworks for solving concrete problems

Content

The following main topics are contained in the course:

feedforward neural networks
recurrent neural networks
convolutional neural networks
backpropagation algorithm
regularization
factor analysis
autoencoders

Literature

See itslearning for syllabus lists and additional literature references.

Examination regulations

Prerequisites for participating in the exam a)

Timing

Autumn

Tests

Project

EKA

N340031112

Censorship

Second examiner: None

Grading

Pass/Fail

Identification

Full name and SDU username

Language

Normally, the same as teaching language

Examination aids

To be announced during the course

ECTS value

Additional information

The prerequisite examination is a prerequisite for participation in exam element a)

Exam element a)

Timing

January

Prerequisites

Type	Prerequisite name	Prerequisite course
Examination part	Prerequisites for participating in the exam a)	N340031101, DM873: Deep Learning

Tests

Oral exam

EKA

N340031102

Censorship

Second examiner: External

Grading

7-point grading scale

Identification

Student Identification Card

Language

Normally, the same as teaching language

Examination aids

To be announced during the course

ECTS value

Indicative number of lessons

48 hours per semester

Teaching Method

At the faculty of science, teaching is organized after the three-phase model ie. intro, training and study phase.

Intro phase (lectures, class lessons) - 32 hours
Training phase: 16 hours, including 16 hours tutorials

The course will consist of frontal lectures supported by discussion sessions. The students will get accompanying exercises demonstrating the collected knowledge on practical real-world problems. The student activation is completed by a mandatory project and discussions of current state-of-the-art research papers during the study phase.

Study activities:

Small take home exercises
Study latest developments and approaches of deep learning by reading recent publications

Teacher responsible

Name	E-mail	Department
Richard Röttger	roettger@imada.sdu.dk	Institut for Matematik og Datalogi, Datalogi, Datavidenskab & Statistik

Timetable

Show full time table

Administrative Unit

Institut for Matematik og Datalogi (datalogi)

Team at Educational Law & Registration

NAT

Recommended course of study

Profile	Programme	Semester	Period