5 ECTS cr / 135 hours of work
|Language of instruction
Autumn / period 2.
Upon completion of the course, the student will be able to
- understand the problem of combining data of different natures and coming from different sources
- explain basic principles of combining multi-sensor data
- know the common types of data fusion techniques
- understand and utilize Bayesian probabilistic reasoning framework in multi-modal data fusion
- understand basic principles of machine learning applied to multi-modal data fusion
- implement basic solutions towards the accomplishment of a given task requiring the integration and combination of data
This course will provide a comprehensive introduction to the concepts and ideas of multi-sensor data fusion. We will be concentrated on defining general statistical framework for multi-modal data processing. Using this framework, we will show concepts of common representation and alignments, sequential Bayesian inference, and machine learning approaches to data fusion as well as specific models and algorithms in each category. Furthermore, the course will illustrate many real-life examples taken from a diverse range of applications to show how they can be benefitted from data fusion approaches.
The course will discuss the following topics:
- Sensors and architectures
- Common representation
- Bayesian inference and probabilistic reasoning
- Sequential Bayesian inference
- Bayesian Decision Theory and ensemble learning
- Advanced topics
|Mode of delivery
The course will be based on a combination of lectures (face-to-face teaching), exercises, and a final project.
|Learning activities and teaching methods
16 h lectures, 16 h exercises (including programming tasks), 35 h final programming project, home study.
The course is suitable for Master level students in Computer science and engineering study programmes, for minor subject studies or for doctoral students.
|Prerequisites and co-requisites
The course will be self-contained as much as possible (i.e., no previous knowledge of multi-sensor data fusion is assumed). Basic knowledge on mathematics and statistics as well as related topics like signal processing, and machine learning will be a plus.
The required prerequisite is the completion of the following courses: 031078P Matrix Algebra, 031021P Probability and Mathematical Statistics, 521156S Towards Data Mining, and 521289S Machine Learning.
|Recommended optional programme components
The course is an independent entity and does not require additional studies carried out at the same time.
|Recommended or required reading
The course will be based on the following textbook: H.B. Mitchell. Data Fusion: Concepts and Ideas. Springer (2012) and selected recent journal articles.
|Assessment methods and criteria
To pass the course, the student should return the exercises, complete a final programming project. Half of the grade will be based on exercises and half on the final project.
The course will utilize a numerical grading scale 1-5. Zero stands for a fail.
Jaakko Suutala and Markus Harju
|Working life cooperation
Course uses Moodle platform.