Laboratory of Feature Engineering and Information Fusion

Recommended Prerequisites

Linear Algebra, Calculus, Statistics.

Teaching Methods

The teaching/learning process observes a collaborative learning approach, which is based on the presentation, analysis, discussion of concepts and techniques in theoretical and theoretical-practical classes.

Theoretical classes: Exposition of concepts, principles, and fundamental techniques in the scope of information fusion. Examples that concretize the practical interest of the subject and exemplify its application to real situations.

Practical classes: Proposed practical problems related to the topics taught in the theoretical classes, their analysis, discussion, and implementation

Learning Outcomes

The discipline of Laboratory of Feature Engineering and Information Fusion aims to provide students with theoretical knowledge and tools to extract, select, merge, and transform information so that it can be efficiently used by algorithms in computational analysis and machine learning.

By the end of the course, the student should be able to:

-Analyze the quality of information (e.g., evaluating the signal-to-noise ratio).

-Extract attributes of different types and domains.

-Develop strategies for data annotation.

-Convert data (e.g., normalization, discretization).

-Fuse information from different sources at three different levels: data, attributes, and decision.

Work Placement(s)

No

Syllabus

Chapter 1: Introduction to Feature Engineering and Information Fusion

 

Chapter 2: Raw Data Processing

-Acquisition and annotation of data

-Quality of data: inconsistency, imperfection, missing data, noise, outliers

-Data preparation: alignment, transformations

-Detection and treatment of nonconformities

 

Chapter 3: Feature Extraction

-Features

-Feature extraction specific to the domain:

--Time Domain

--Frequency Domain

--Time/Frequency Domain

--Nonlinear Domain

 

Chapter 4: Data Fusion

-Data-level fusion: weighted average, (Extended) Kalman Filter, particle filtering

-Feature-level fusion: parametric and non-parametric methods

-Decision-level fusion: classical and Bayesian inference, Dempster-Shafer theory, Markov Monte Carlo models

-Model combination: voting, adaboost, random forest.

Assessment Methods

Assessment
Exam: 50.0%
Project: 50.0%

Bibliography

Additional References:

•          Linear Models and Time-Series Analysis Regression, ANOVA, ARMA and GARCH; Marc S. Paolella; 2019 JohnWiley & Sons Ltd; ISBN 9781119431985 (ePub) | ISBN 9781119431909

•          Time-Series Prediction and Applications - A Machine Intelligence Approach; Amit Konar, Diptendu Bhattacharya; 2017 Springer; ISBN 978-3-319-54596-7

•          Elements of Nonlinear Time Series Analysis and Forecasting; J. De Gooijer; 2017 Springer Series in Statistics; ISBN 978-3-319-43251-9

•          Nonlinear time Series analysis; Ruey S. Tsay, Rong Chen; 2019 John Wiley & Sons, Inc.; ISBN 978