Filling the gap between theory and practice in Machine Learning: an engineer’s perspective
Event date:  January 28, 2019 4:00 PM  to 5:00 PM

Since the recent (re)discover and great success of Deep Neural Networks it has become clear that Machine Learning has been developing on two parallel and somewhat not-so-converging paths. On one hand theoretical results achieved by Statistical Learning Theory (SLT) allow to derive new learning approaches and build a theoretical framework for answering questions like “how can we select the optimal predictive model?”, “how can we rigorously assess its prediction performance?”.  This is of paramount importance in industrial and engineering applications where the risk of using a ML algorithm, which learns from empirical data, should be estimated with reasonably accuracy. On the other hand, theory seems of limited applicability to practical problems, because learning algorithms achieve incredibly successful results, in many applications, that SLT cannot explain. 

In this seminar I will summarize recent theoretical advances in this direction and some engineering applications of ML algorithms.

 

Biography: 

Davide Anguita (MSc - 1989,  PhD - 1993) is Full Professor of Information Processing Systems at the University of Genova, Italy (Department of Informatics, Bioengineering, Robotics and Systems Engineering), where he teaches “Business Intelligence” (Industrial Eng.), “Data Analysis and Data Mining” (Computer Eng.) and “Theory and Practice of Learning from Data” (Doctorate in Computer Science), after holding visiting positions at the Hewlett-Packard Laboratories (HP-Labs, Analytical/Medical Dept., Palo Alto, CA, USA), the International Computer Science Institute, Berkeley, USA and the University of Trento, Italy.

His main research interests are in the field of theory, methodologies and industrial applications of Machine Learning and Data Analysis. He has been principal investigator of numerous research and technology transfer projects (Finmeccanica, National Institute for Cancer Research, Whirlpool Europe, etc.) and was in charge of the research agreement between the University of Genova and Ferrari S.p.A. for the application of intelligent data analysis to FIA Formula One World Championship (F1) auto racing.

He has been Member of the EC-Network of Excellence NeuroNet 1 and NeuroNet 2, Chair of the Smart Adaptive Systems section of the EC-Network of Excellence EUNITE (European Network of Intelligent Technologies for Smart Adaptive Systems), and Chair of the EC Concerted Action NiSIS (Nature-inspired Smart information Systems) Focus Group in “Data Technologies”. He has been principal investigator for the University of Genova of the EC-H2020 project In2Rail (Innovative Intelligent Rail – coordinator: Network Rail, UK) and currently of the EC-H2020 project IN2DREAMS (INtelligent solutions 2ward the Development of Railway Energy and Asset Management Systems in Europe). He has been evaluator of European (EC-FP7 and H2020), national (MIUR) and regional (Liguria, Veneto) research projects.

He co-founded "Smartware & Data Mining S.r.l.", “Novigo Technology S.r.l.” and “Zenabyte S.r.l.”, spin-offs of the University of Genova, operating in the field of Business Intelligence and Industrial Analytics.

He is co-author of more than 170 publications on international journals or refereed conference proceedings, two patents on intelligent data analysis methods, and is a Senior Member of the IEEE.

URL:
https://www.dibris.unige.it/en/anguita-davide
http://www.smartlab.ws/

  

  

  

Location

University of Konstanz, Universitätsstr. 10, Konstanz
Powerwall C202

University of Stuttgart, VISUS-Building, Allmandring 19, Vaihingen
Powerwall Room -01.116 (Live Transmission)