COURSE SYLLABUS

Neural networks and learning machines 7.5 credits

Second cycle, D7046E
Version
Course syllabus valid: Spring 2021 Sp 3 - Present
The version indicates the term and period for which this course syllabus is valid. The most recent version of the course syllabus is shown first.

Syllabus established
by Jonny Johansson, HUL SRT 15 Feb 2019

Last revised
by Jonny Johansson, HUL SRT 21 Feb 2020

 Education level Second cycle Grade scale G U 3 4 5 Subject Computer Science and Engineering Subject group (SCB) Computer Technology Main field of study Computer Science and Engineering

Entry requirements

Good understanding of mathematical analysis and linear algebra; ability to develop and apply computer programs to solve mathematically formulated problems, eg D0009E Introduction to Programming or D0017E Introduction to Programming for Engineers; understanding of basic mathematical statistics including probability distributions, expectation and variance, such as S0001M Mathematical Statistics or S0008M Probability Theory and Statistics; understanding of basic signal processing including sampling, time-discrete processing of time-continuous signals, linear and time-invariant systems, such as S0001E Signal Analysis or S0004E Signals and Systems; understanding of basic electrical theory including RC circuits, eg E0003E Electric Circuit Theory or E0013E Fundamentals of Electrical Engineering. More information about English language requirements [http://www.ltu.se/edu/bli-student/Application-process/English-language-requirements-1.109316?l=en].

Selection

The selection is based on 20-285 credits

Course Aim

After completion of the course, the student should be able to:

• Describe and differentiate artificial intelligence, machine learning, artificial neural networks, and neuromorphic engineering.
• Describe the function of formal neuron models and neural network architectures, including spiking neural networks, feedforward networks, convolutional networks and recurrent networks.
• Demonstrate how neural networks can be trained, validated and tested with supervised and unsupervised methods and analyse how different hyperparameters and regularization affect model generalisation.
• Evaluate strengths and weaknesses of neural network models and make comparisons with other machine learning methods like linear and logistic regression, including the aspects of computing requirements, data requirements, bias, variability, sensitivity and generalisation.
• Give examples of learning machines involving neural network processors and neuromorphic systems and describe their properties and motivation, including ethical considerations.
• Develop and evaluate a neural network model that address a particular engineering problem, which involves for example pattern recognition, clustering, regression, anomaly detection, recommender systems, or reinforcement learning.
• Formulate and implement derivations of basic neural network concepts introduced in the course material and assess the validity of the results.

Contents

Neural network models and processors are used to solve pattern recognition, data modelling and automatic control problems that are difficult to address with other modelling approaches. This course introduces concepts and methods required to design, train and validate neural networks, and to determine when the use of such models is motivated. Furthermore, similarities and differences between artificial and biological neural networks are described. The course introduces concepts and methods needed for further studies of advanced neural network topics and implementations for learning machines and neuromorphic systems.

Realization
Lectures, read and view self-study material, laboratory work in the form of simulation exercises, self-assessment of learning with peer-review, and project work with oral presentation. Projects can include industrially related problems and datasets.

Examination
Assessment of laboratory work and project work. Written self-assessment. Oral examination.

Remarks
The laboratory exercises require basic knowledge of Python. Guidance for preparatory self-studies are available in the course room.

Examiner
Fredrik Sandin

Literature. Valid from Autumn 2019 Sp 1 (May change until 10 weeks before course start)
Deep Learning. I. Goodfellow, Y. Bengio and A. Courville, MIT Press, 2016;
http://www.deeplearningbook.org

Neuronal Dynamics: From single neurons to networks and models of cognition.
W. Gerstner, W. M. Kistler, R. Naud and L. Paninski, Cambridge University Press, 2014;
http://neuronaldynamics.epfl.ch

Course offered by
Department of Computer Science, Electrical and Space Engineering

Modules