September 2nd - 5th, 2016, Robert Gordon University, Aberdeen
The proceedings will be published in the Springer series
Communications in Computer and Information Science

Machine Learning and Nature Inspired Techniques in Industry 4.0

Maurizio Fiasché, Ph.D., SIEEE, Politecnico di Milano,
Industrial Relations Co-Chair of the IEEE Italy Section,
SIG EANN member,

Find more about this tutorial

Generative and Adversarial Deep Learning

Amos Storkey and Harri Edwards, University of Edinburgh Research Group Information

This tutorial will review modern developments in deep learning beyond the simple business of predicting a class label. After a quick review of the basics of deep learning, its implementation and the core reasons why deep learning methods have proven to provide the performance advantages they have, we will explain how models for multiple predictors can be developed that capture the correlations between those variables. We will introduce the use of adversarial methods for avoiding the difficulties of intractable normalisation constants. Finally we will show a number of exciting recent directions and engineering examples and applications of

Variable selection for efficient design of neural networks and other machine learning-based models: efficient approaches for industrial applications

Valentina Colla, Pd.D., Silvia Cateni M.Sc., Scuola Superiore Sant'Anna, Pisa, Italy
Find more about this tutorial

In the tutorial, different categories of variable section procedures will be presented and discussed, by also proposing exemplar case studies coming mainly from the industrial field, in order to highlight strengths and weaknesses of each method in relation to the different tasks and to the variables of the considered dataset. Finally some open challenges and future directions in the research on this field will be outlined.

Classification of unbalanced datasets for the detection of rare patterns: stateof the art and current challenges with a special focus on industrial problems

Valentina Colla, Pd.D., Marco Vannucci, Ph. D., Scuola Superiore Sant'Anna, Pisa, Italy
Find more about this tutorial

In the proposed tutorial the main issues related to the classification of unbalanced datasets will be described and the main factors that affect the performance of standard classifier will be analysed. Subsequently the state-of-the-art of the approaches for coping with this problems will be presented including internal, external and hybrid methods (i.e. a combination of internal and external techniques). Particular attention will be given to emerging methods based on the use of Artificial Neural Networks (ANN) which, due to their characteristics of flexibility, robustness and generalization capabilities, are becoming the leading technology in this framework.