Swedish web site
Deep Learning2025VT, 6.0 creditsFull
|
|
Course plan
Lectures
Tentative lecture plan:
- Introduction
- Optimization, regularization
- Representation learning in vision I
- Generative models I
- Generative models II
- Sequence models
- Transformers
- Representation learning in vision II
- Graph neural networks
- Uncertainty quantification
- Summary
Recommended for
The course was last given
VT2024 (as MSc course)
Goals
The overall aim of the course is to give the student a good theoretical
understanding and good practical skills in machine learning based on deep
neural networks. On completion of the course, the student should be able to:
1. Use central theoretical concepts to describe and account for differences
between models and methods in deep learning
2. Use relevant models and methods from deep learning to formulate, structure
and solve practical problems involving large and complex data sets
3. Choose appropriate evaluation protocols and critically evaluate the quality
of the solutions
4. Analyze and criticize the choice of models and methods
5. Describe ethical issues and societal aspects related to deep learning.
Prerequisites
Calculus in one and several variables, linear algebra, statistics, optimization and programming. Basic machine learning (at least 6 credits) including artificial neural networks. Programming in Python and experience in programming for numerical problem solving and linear algebra (e.g. via NumPy, Matlab or Julia).
Contents
The course introduces key concepts in deep learning, from basic to
state-of-the-art.
This includes: Common model types (e.g., MLP, CNN, RNN, Transformers);
Optimization and learning algorithms for deep networks; Representational
learning and self-supervised learning; Generative models; Different application
areas of deep learning, in e.g. natural language processing and text analysis,
computer vision, and natural sciences; Ethical considerations and societal
aspects linked to deep learning.
Organization
Literature
Main reference:
Bishop and Bishop, Deep Learning: Foundations and Concepts, Springer 2024.
https://www.bishopbook.com/
Lecturers
Examiner
Fredrik Lindsten
Examination
This course is given together with the MSc level course TDDE70. However, the
examination differs a bit between the MSc and PhD courses. To pass the course
as a PhD students you need to:
- Pass all labs (same as MSc students), but the labs are solved independently
without TA support (individually or in pairs)
- Define and carry out a course project (individually or in small groups) +
write a project report.
- Get a "pass with distinction" grade on the exam.
Credits
6
Organized by
STIMA, AIICS, CVL
Comments
This is an MSc level course. PhD students attending the course are expected to work more independently and may not get the same level of TA support as the MSc students. PhD students should solve the labs on their own.
Page responsible: Anne Moe
