This translation is for information purposes only. In the event of discrepancies, the Swedish-language version takes precedence.

Deep Learning, 7.5 Higher education credits

Djupinlärning, 7,5 Högskolepoäng

Established: 2022-09-01

Established by: Department of Engineering Science

Applies from: V23

Learning outcomes

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

Knowledge and understanding

• show understanding of the way artificial neural networks (ANN) and Deep Neural Network (DNN) get trained.
• explain the signal flow in an ANN.
• discuss the workings of some of the modern ANN and DNN architectures, e.g. CNN, LSTM, GAN etc.
• demonstrate knowledge and understanding of the "no free lunch theorem" in the sense of appreciating the limitations and strengths of various ANN/DNN architectures.

Competence and skills

• select the right approach to solve a practical problem using ANN/DNN.
• utilise existing libraries to build ANN/DNN applications.
• design a pertinent set of performance matrices based on which they can validate the performance of their algorithm(s).
• evaluate the performance of algorithms based on a pragmatic set of performance matrices.
• use the results from the experiments to re-design and finetune the ANN/DNN architectures so as to meet the requirements of the task.

Judgement and approach

• relate to some of the salient aspects of the EU AI Regulation.

Entry requirements

General entry requirements and approved result from the following course/courses:

IAI600-Introduction to Artificial Intelligence and Machine Learning and

BSD600-Big Data Processing and Analysis and

PFA600-Programming for Automation and

STB600-Sensor Technology and Image Analysis or the equivalent.

The forms of assessment of student performance

Written laboratory-work report in group. Written project report in group with oral presentation.

Course contents

This course shall introduce the students to the modern algorithms around deep learning. It will also expose students to practical applications. The following are the main contents of this course:

• Revision of linear algebra and calculus necessary to understand ANN and DNN.
• Revision of generic AI principles (e.g. bias-variance trade-off, meaning of validation vs test etc.).
• Perceptron and back-propagation algorithm.
• Deep learning networks.
• Some interesting DNN networks (e.g. CNN, LSTM etc.)
• Applications in some domains (e.g. computer vision, NLP etc.).
• Some modern DNN architectures (e.g. encoder-decoder, GAN etc.).
• Practical aspects of applying DNN/ANN in a real-life challenge.

Other regulations

Course grading: F/Fx/E/D/C/B/A - Insufficient, Insufficient- more work required before the credit can be awarded, Sufficient, Satisfactory, Good, Very Good, Excellent

Course language: The teaching is conducted in English.

General rules pertaining to examination at University West are available at www.hv.se.

If the student has a decision/recommendation on special support due to disability, the examiner has the right to examine the student in a customized examination form.

Cycle

Second cycle

Progressive specialization

A1F - second cycle, has second-cycle course/s as entry requirements

Main field of study

Automation, Computer Engineering, Computer Engineering