SSIE 2018 is dedicated to highly energy efficient and sustainable ICT, both from a technological viewpoint (innovative microelectronic devices) and from the point of view of communication networks (energy harvesting mobile networks). This edition of the summer school was conceived and structured as a dissemination vehicle for two European projects, as follows.
Project SCAVENGE, with prof. Michele Rossi as Prime Investigator, centered on the design of architectures and algorithms highly energy efficient 5G cellular networks, able to harness the energy from renewable sources (e.g., solar and wind).
Efficient power conversion systems are at the heart of the worldwide effort for a green economy, since they can minimize losses and save energy. Semiconductor power devices are a central part of any power conversion circuit and are ubiquitous in our daily lives: they transform voltages for a multitude of appliances, such as from the 220V AC mains to a 12V DC end-user appliance and enable to convert from DC (such as a battery in an electric car) to AC (such as a motor drive) and vice versa. Highly efficient power switching devices are a key for successful introduction of full electric vehicles into the market. The InRel-NPower project aims to contribute to this world-wide challenge through the development of GaN- and AlN-based power devices.
The fifth generation 5G of mobile technology will support 1,000 times more capacity per unit area than 4G, for more than 100 billion devices with typical user rates of 10 Gb/s, and significantly lower latency and higher reliability. The higher capacity demanding human-centric communications will be complemented by an enormous increase in the number of communicating machines. However, this enormous growth in the number of devices and access points will also lead to an equally large growth in the carbon footprint of the information and communication technologies (ICTs). Connecting this dense network of BSs to the energy grid, and regularly recharging drained end device batteries is physically impractical, if not impossible. SCAVENGE tackles sustainable design, protocols, architectures and algorithms for next generation 5G cellular networks. Our overall purpose is to allow mobile systems and especially their constituting base stations, femto, small-cells, mobile devices and sensors to take advantage of sources harvesting ambient energy (such as renewable sources).
8:45 – 9:00 Gaudenzio Meneghesso, Michele Rossi Department of Information Engineering, University of Padova: “Summer School introduction, program description, introduction to InRel-NPower and SCAVENGE”
9:00 – 11:00 Plenary Florent Cadoux, Fondation Partenariale de Grenoble INP / laboratoire G2Elab, “Distributed Control of Smart Power Grids: an Industrial Viewpoint”
11:00 – 13:00 Plenary Florent Cadoux, Fondation Partenariale de Grenoble INP / laboratoire G2Elab, “Distributed Control of Smart Power Grids: Selected Problems”
14:00 – 16:30 Plenary Peter Moens, On Semiconductor, Belgium “General Overview of GaN based Power Devices” Download-Moens_presentation
9:00 – 10:30 Paolo Dini, Centre Tecnològic de Telecomunicacions de Catalunya (CTTC), Spain, Network “Energy Sustainable Architecture and Methods for Future Mobile Communication Networks”
11:00 – 12:30 Lorenza Giupponi, Centre Tecnològic de Telecomunicacions de Catalunya (CTTC), Spain, Network “Machine Learning Meets Self-Organizing Networks”
14:00 – 15:00 J. Derluyn, EpiGaN, Belgium, Technology “Materials and Epigrowths for wide bandgap semiconductors devices” Download_Derluyn_presentation
15:00 – 16:00 E. Meissner, Fraunhofer IISB, Germany, Technology “Defects in Nitride Semiconductors Materials and their relevance to electronic devices” Download_Meissner_presentation
16:15 – 17:30 F. Medjdoub CNRS-IEMN Lille, France; Technology “Device processing and architectures in Gallium Nitride based semiconductors” Download_Medjdoub_presentation
9:00 – 10:30 G. Meneghesso, Department of Information Engineering, University of Padova, Technology “Reliability of GaN-power transistors: an overview” Download_Meneghesso_presentation
11:00 – 12:30 M. Rittner Robert Bosch GmbH, Germany; Technology “Modern high Performance Assembly and Interconnection Technologies for Wide-Band-Gap Power Semiconductor basing Power Electronics” Download_Rittner_presentation
10:00 – 11:00 (*) Paolo Tenti, Dept. of Information Engineering, Univ. of Padova; Network “From Micro-Grid to Energy LAN to Internet of Energy: a pathway to electric revolution – Part I”
11:30 – 12:30 (*) Paolo Tenti, Dept of Information Engineering, Univ. of Padova; Network “From Micro-Grid to Energy LAN to Internet of Energy: a pathway to electric revolution – Part II””
14:00 – 15:30 Fabio Aiolli, Department of Mathematics, University of Padova; Network “Representation and Kernel Learning – Part I”
16:00 – 17:30 Fabio Aiolli, Department of Mathematics, University of Padova; Network “Representation and Kernel Learning – Part II”
19:30 – 22:00 Social Dinner
09:00 – 10:30 Andrea Passerini, Department of Information Engineering and Computer Science, University of Trento; Network “Inference and Learning with Bayesian Networks – Part I”
11:00 – 12:30 Andrea Passerini, Department of Information Engineering and Computer Science, University of Trento; Network “Inference and Learning with Bayesian Networks – Part II” (video)
14:00 – 15:30 K. Kriegel, SIEMENS Germany, Technology “Industrial applications of wide bandgap semiconductors devices”
16:00 – 17:30 P. Bleus CE+T, Belgium; Technology “Applications in Energy saving of wide bandgap semiconductors devices”
9:00 – 12:00 Plenary A. Testolin, Department of General Psychology, University of Padova. “Discovering latent structure from big data using neural networks”
12:00 – 12:30 School closing
Abstract: The course will start with some basics about optimization in general, from the theoretical viewpoint (mathematical notions, etc) as well as from the practical viewpoint (modelling languages, solvers, etc). We will then review some of the classical applications of optimization in the power industry today. Then we will turn to more prospective applications of optimization to future “smart grids”
Bio: Florent Cadoux studied at Ecole Polytechnique (France) and holds a PhD in Applied Mathematics and Optimization from INRIA (2009), where he worked on numerical optimization and convex analysis with various applications to combinatorial optimization, computational mechanics, and computer graphics. He has also been a consultant in optimization at Artelys, where he worked on various applications of optimization to the electric industry. He currently holds the “Enedis industrial chair on Smartgrids” at Fondation Grenoble-INP / G2Elab (Grenoble Electrical Engineering Laboratory).
Abstract: this course will provide an introduction to GaN based power devices. The following topics will be addressed (each topic including state-of-the art literature review):
1) Why a new generation of more efficient power devices is needed?
2) Derive the Baliga Figure of Merit for power devices (student exercise) and interpretation of the result. Challenge the assumptions which are used to derive the equation.
3) Brief overview of the main materials for power devices (Si, SiC, GaN, AlN, Ga2O3 and diamond);
4) Basic material properties of GaN;
5) Polar versus non-polar materials.
6) Concept of polarization charge;
7) Ternary and quaternary alloys;
8) Different device architectures: High Electron Mobility Transistors (“HEMT”) versus standard junction transistors: impact of device concept on on-resistance and capacitance (make link to the Baliga figure-of-merit);
9) Depletion mode and enhancement mode devices. Basic band structure;
10) Lateral and vertical GaN-based devices;
11) Dynamic Ron, and how to characterize;
12) Application testing results.
Bio: Peter Moens received a M.Sc. and a Ph.D. in solid state physics from the University of Gent, Belgium, in 1990 and 1993 respectively. From 1993 till 1996, he worked as a post-doctoral fellow in collaboration with Agfa-Gevaert, Mortsel Belgium. In 1996, he joined ON Semiconductor, Oudenaarde, Belgium where he was involved in the technology and device development for smart power applications, and the related reliability aspects. Since 2008 he is responsible for the development of 600+V discrete power devices, both in silicon as well as in wide band gap materials. He is/was a member of the technical program committees of IEDM, ISPSD, IRPS, IRW, ESSDERC and ESDEOS Symposium. He was Vice-chair of the integrated power subcommittee of IRPS 2005 and 2008, and Chair of the same committee of IRPS 2006 and IRPS 2007. He was the Technical program chair of ISPSD 2009, and the General chair of ISPSD 2012. He was the subcommittee chair of the Power and Compound semiconductor devices subcommittee for IEDM 2014. He authored and co-authored over 150 publications in peer reviewed journals or conferences, of which over 60 as first author. He authored or co-authored of 12 invited papers and is the recipient of 3 best paper awards. He also presented tutorials on smart power reliability at IRPS 2005, IRPS 2006 and ISPSD 2007, and on GaN power device reliability at IRPS2015. He holds over 25 patents.
Abstract: We are now living the digital era. Dematerialization is becoming a reality, and everybody and everything, including machines, is globally connected through the Internet. The trend is of a further increase in traffic demand, number of offered services and connected devices, especially mobile. However, the massive use of Information and Communication Technologies (ICT) is also increasing the level of energy consumed by that system and its footprint on the environment. In 2030 ICT is expected to consume 51% of the electricity generated and will be responsible of 23% of the carbon footprint by human activity. Sustainable design of ICT, and specially of mobile networks, is, therefore, a key and challenging sector for societal prosperity. In this talk, we will elaborate on the architecture of the future mobile networks, often referred to as 5G (5th Generation), and its interaction with the electricity grid. The integration of the radio access network with a distributed renewable energy system will be discussed, by reporting the main building blocks and methods to achieve the self-sustainability of the integrated system.
Bio: Paolo Dini (MSc 2001, PhD 2005) is a Senior Researcher at the Communication Networks division of the Centre Tecnologic de Telecomunicacions de Catalunya (CTTC). Formerly he worked as a research assistant in the Computer Science and Systems (DIS) and Information and Communication (INFOCOM) departments of the Università di Roma “La Sapienza” (2001 – 2005). In 2005, he worked with the Research Center on Software Technologies (RCOST) as a research assistant. In the same year, he was also an assistant professor at INFOCOM department lecturing the course of “Comunicazioni Elettriche”. In 2008 and 2011 he received two grants from the Cisco Research Centre (San José, US) for his research on heterogeneous mobile networks. Paolo has an active participation in research projects and he is currently the coordinator of the SCAVENGE European Training Network on energy harvesting cellular networks. He published more than 50 papers in scientific journals, magazines and international conferences and serves many IEEE conferences and journals as a technical reviewer. His research interests encompass wireless networks modelling and optimization, with particular emphasis on energy saving and energy efficiency, energy sustainable network architectures, protocols and algorithms, smart grids, self-organizing networks, cognitive networking and machine learning, radio resource and mobility management.
Abstract: in this talk, after a brief overview and taxonomy of the main machine learning tools and self-organizing networks (SON) concepts, with special attention to energy efficiency, we will show how machine learning can be a powerful instrument to implement self-organization in sustainable 4G/5G networks. We argue that currently mobile networks are generating a huge amount of data in the form of network measurements, network control and management interactions, which can be properly exploited to better manage the network through machine learning approaches. 5G is expected to make this amount of data even bigger. We will identify the different categories of useful information that can be extracted from a 4G network and show how this information can be useful to better address multiple SON use cases, giving specific emphasis to energy saving use case. We will discuss the approach that we are following to extract data and measurements from real networks, to analyze it and make control decisions based on that. Also, we will show that reinforcement learning, and in particular Time Difference learning, can be an effective tool to online address energy saving problems, in both centralized and distributed cellular architectures and for both macro/small cell deployments. We will describe different examples where SON functions benefit from the experience gained through machine learning. Finally, we will highlight challenges and directions for future works.
Bio: Lorenza Giupponi (MSc 2002, PhD 2007) received the Telecommunications Engineering degree from University of Rome “La Sapienza” in July 2002 and the PhD from the Technical University of Catalonia (UPC) in 2007. She joined the Radio Communications Group of UPC in 2003 with a grant of the Spanish Ministry of Education, in the context of the “Formación Profesorado Universitario” program. During 2006 and 2007 she was assistant professor in UPC. In September 2007 she joined the CTTC where she is currently a Senior Researcher in the Mobile Networks Department of the Communication Networks Division. Her research experience falls in the area of radio resource management, interference management, small cells, self-organizing networks, LTE, LTE-A, LTE in unlicensed, heterogeneous networks, mmWave, simulation modelling. Her technical experience is enriched by a wide management experience. Since 2007 she is also a voting member of the Executive Committee of CTTC, where she acts as the Director of Institutional Relations. She is the co-recipient of the IEEE Consumer Communications and Networking Conference 2010 (IEEE CCNC 2010) and of the IEEE third International workshop on Indoor and Outdoor Femto Cells 2011 best paper awards. Two of her Transactions journals have been selected by IEEE ComSoc among the BEST READINGS in the areas of Radio Resource Management and Cognitive Radio, respectively. Since 2015 she is a member of the Executive Committee of ns3 consortium and she leads the activities of CTTC in the context of ns-3, as main developers and maintainers of the LTE module. She has led and participated in multiple EU (in FP6, FP7 and H2020, 5GPPP), national (she was the Principal Investigator of two projects funded by the Spanish national research program) and industrial projects (she was the Principal Investigator of multiple projects funded by the Qatar National Research Fund, the WiFi Alliance and Silicon Valley based startups active in the areas of small cells, SON and unlicensed LTE). She is an IEEE Senior Member.
Abstract: This 1h course will provide an introduction to material aspects of GaN and its epitaxy. The following topics will be addressed:
1) III-V semiconductor overview (bandgap vs. lattice constant),
2) Device examples (build an LED);
3) Polarity in crystals,
5) Derive 2DEG from polarization charges;
6) HEMT / Surface passivation;
7) Epitaxial growth;
8) MOCVD and other techniques;
9) Basic concepts (Boundary layer model, thermodynamics);
10: Material characterisation techniques; 1
12) GaN hetero-epitaxy, esp. GaN on Si;
14) Buffer concepts.
Bio: Joff Derluyn is a co-founder and the CTO of EpiGaN (www.epigan.com). In that capacity, he is responsible for the R&D roadmap, IP strategy and main responsible for bilateral and public funded projects (ESA, EU-FP7, EU-H2020, EU-ECSEL). Joff has over 20 years of experience in MOCVD epitaxy and device processing of III-N and III-V semiconductor materials and devices, holds over 25 granted patents in the field of III-N electronics and is author or co-author of over 100 peer-reviewed publications. He acts as reviewer for high impact scientific journals such as APL, EDL and TED and is a member of NEREID expert committee to set up a roadmap for wide bandgap semiconductors. Prior to founding EpiGaN, Joff was at imec as the responsible for the GaN devices and characterization team. He holds a PhD in electrical engineering from Ghent university on the topic of epitaxial growth of dilute nitrides.
Abstract: This short lecture will provide information with regard to the defect microstructure of nitride semiconductors and the implication for the electrical performance of those materials. The following topics will be covered: 1) Short introduction to the technology of nitride semiconductor crystal growth, 2) The real structure of semiconductor crystals, 3) Problems associated to hetero-epitaxy, especially resulting extended defects, 4) Properties of dislocations, 5) Why are dislocations relevant to an electronic device?, 6) Why we would need native nitride substrates.
Bio: Elke Meissner is a senior scientist at the Department Materials of the Fraunhofer Institute of Integrated Systems and Device Technology (IISB) in Erlangen, Germany and head of the Nitrides group which operates at three locations. Besides the IISB she manages GaN activities at the Frauhofer Technology Center for Semiconductor Materials (THM) at Freiberg (Saxony, Germany) and the Technical Faculty of University of Erlangen-Nuremberg at the chair for Electron Devices, where she also holds a lecture ship. She received her PhD from the University of Bayreuth, Germany in 2000. After some time spent at the German Aerospace Research Centre, she moved to other research institutions and universities to further work with crystalline matter. Elke now offers a broad background in process-properties-correlation of advanced materials from more than 20 years of experience in various fields ranging from applied & experimental mineralogy, Si3N4 and related processes under high pressure and temperature as well as the crystal growth of GaN an AlN. Her recent work is strongly focused on the structural characterization of crystals and crystal layers of novel semiconductor materials and the influence of structural disturbances on electrical reliability, especially of electron devices based on nitrides. She is inventor or co-inventor of international patents and authored or co-authored numerous scientific publications, contributions to journals and conference, has been chair or co-chair of international conferences, reviewer of many recommended scientific journals and is member of the international steering committee of the International Workshop on Bulk Nitride Semiconductors (IWBNS).
Abstract: Group-III nitride, gallium nitride (GaN) is a promising wide-bandgap semiconductor. GaN-based High Electron Mobility Transistors were originally developed for high-power high-frequency amplifiers, however, in the last decade research and dedicated development efforts were carried out on AlGaN/GaN HEMTs grown on large diameter silicon substrates producing cost-effective high-efficiency power switching devices, which commercial companies have started to mass produce. They have a great impact on consumer electronics, permitting advanced power electronics due to their excellent performance such as low on-resistance and high switching speed. In this lecture, an in-depth description of the device processing and architectures in Gallium Nitride based semiconductors will be given.
Bio: Farid Medjdoub is a CNRS senior scientist and head of the wide bandgap activities at IEMN in France since 2011. He received his Ph.D. in Electrical engineering from the University of Lille in 2004. Then, he moved to the University of Ulm in Germany as research associate before joining IMEC as a senior scientist in 2008. Multiple state-of-the-art results have been realized in the frame of his work. Among others, world record thermal stability up to 1000°C for a field effect transistor, best cut-off frequency / breakdown voltage or highest lateral GaN-on-silicon breakdown voltage using a local substrate removal have been achieved. His research interests are the design, the fabrication, and characterization of innovative GaN-based devices. He is author and co-author of more than 100 papers in this field. He holds several patents deriving from his research. He serves as a reviewer for IEEE journals and is a TPC member in several conferences. He is also part of the French observatory of wide bandgap devices
Abstract: The past few years have been exciting and extremely productive for the GaN community, and the research in the field of GaN-power devices has shown impressive advancements. In lateral GaN HEMTs, a two-dimensional electron gas (2DEG) is formed at the interface between GaN and AlGaN; the high mobility of the 2DEG (in excess of 2000 cm2/Vs) results in current densities around 1 A/mm, and in a very low on resistance (25 mΩ for a 650 V/60 A device). This implies a significant reduction in the resistive and switching losses, compared to silicon devices, and this has a positive impact on the efficiency of GaN-based power converters (kW-range power converters with efficiency higher than 99% have already been demonstrated, based on GaN HEMTs). A relevant aspect that is currently under study is the reliability of GaN-based transistors. In fact, during operation in high-voltage power converters, the HEMTs may be subject to extreme field and current levels that may favor device degradation. In real-life applications, several potentially harmful conditions may be reached, favoring the exposure to off-state bias, semi-on stress conditions, hard switching, and high gate bias. This presentation reviews the most recent results published in the field of GaN-based power transistors. After a brief introduction, we discuss the latest technology developments that have an impact on device reliability. Then, we summarize the most important issues related to the degradation of GaN-based HEMTs submitted to stress regimes (off-state, hard switching, positive gate bias, high temperature, …).
Bio: Matteo Meneghini received his PhD in Electronic and Telecommunication Engineering (University of Padova), working on the optimization of GaN-based LED and laser structures. He is now assistant professor at the Department of Information Engineering at the University of Padova. His main interest is the characterization, reliability and simulation of compound semiconductor devices (LEDs, Laser diodes, HEMTs). Within these activities, he has published more than 300 journal and conference proceedings papers. During his activity, he has cooperated and/or co-published with a number of semiconductor companies and research centers including: -OSRAM-OptoSemiconductor (Germany), -Panasonic Corporation (Japan), Universal Display Corporation (USA), -NXP (The Netherlands), -ON Semiconductors (Belgium/USA), -Sensor Electronic Technologies (USA), -IMEC (Belgium), -Infineon (Austria), -Fraunhofer IAF (Germany), -University of Cambridge (UK), -Universiy of California at Santa Barbara (USA), -University of Wien (Austria). Meneghini is a Senior Member of IEEE and a member of the SPIE. He – together with his colleagues – won several best paper awards at international conferences (including ESREF 2009, IWN 2012, ESREF 2012, ESSDERC 2013). He has/is participated to the technical committee of several conferences including IEEE-IEDM, IEEE-IRPS, ESREF, ESSDERC.
Abstract: The usage of wide-band-gap-basing (e.g. like GaN power semiconductors) power electronics, offers a huge potential for further increase of converting efficiencies, higher power densities and the ability of mechatronic integration in nearly all applications and aggregates. But in parallel there is a decisive demand for high performance assembly and interconnection technologies in the means of high temperature capability, high power cycle robustness, low inductive design elements and integration ability of further active and passive components (capacitors, inductors, sensors, driver & controller ASICs) in the nearest ambient of the power chip. Exclusively with new assembly and interconnection technologies and design elements the outperformance of the new WBG-basing power electronics’ systems – in comparison to state of the art Si-basing power electronics – will succeed. Beginning with an overview of modern power electronics substrate and circuit carrier technologies the mounting and contacting of the power semiconductors on those with robust technologies, like the so-called Silver-sintering and diffusion soldering technology or the Cu-bonding technology, will be described. Further on necessary power stage design elements regarding fast and hard switching of the power switches will be described. The functional characteristics of the commutation cell needs the completion by the DC-link capacitor. Therefor a summary of usable capacitor technology and the boundaries regarding its integration in the commutation cell will be described. A lot of industrial branches need special robustness capabilities of their power electronics’ systems. These are typically limited by the chosen hardware technologies and the involved materials. Specific ambient loads like certain temperatures, vibration, chemical substances lead to ageing and degradation of power electronics hardware. And the functional use itself induces ageing conditions as well. Applying the on-current and switching it off leads to a specific load on the assembly and interconnection technology and the module’s material stack. A survey on the power electronics’ robustness issue will be given.
Bio: Dr. Martin Rittner is senior expert for power electronics assembly and interconnection technologies in the Corporate Research unit of Bosch. He studied Physics at the University of Stuttgart and received his diploma in Nuclear Physics in 1994. Afterwards he did his PhD thesis in the field of Semiconductor Physics at the same university. Since working as research employee in the Corporate Research sector of Bosch in 2001 he attended several German and EU public funded projects in the field of electronics packaging and assembly technologies for automotive and power electronics applications. For the German automotive supplier industry, he is currently the head of the ZVEI-ECPE working group ‘High Temperature and Power Electronics’. In 2015, he rounded his academic skills by finalizing his economic studies and receiving the degree as Master of Business Administrations (MBA).
Abstract: Owing to the increasing penetration of information and communication technologies and power electronics, low-voltage microgrids with distributed generation will progressively evolve to Energy LANs (E-LANs), that is, energy networks with degrees of flexibility, reliability, robustness, and readiness similar to LANs of digital devices. This will result in a full and synergistic exploitation of the control potential of distributed energy sources, driven by advanced grid supervisors targeting both microgrid performance and demand response at utility terminals. In addition, prosumers will gain the capability to manage and share their energy resources to improve power quality at their premises, save energy and money, and act as individual, or even aggregate, energy traders. This speech introduces an optimum control approach for E-LANs that relates to the general case of poly-phase meshed grids with distributed energy sources and multiple connections to the utilities. Thanks to a synergistic control of distributed grid-tied power inverters, the proposed approach assures efficiency, scalability, plug-and-play integration of energy resources, quick response to power demands, and prevents discontinuity of operation in case of fault. In addition, it makes possible prosumers’ aggregation to form communities of real-time energy traders, thus moving a step ahead toward E-LANs and to the future Internet of Energy.
Bio: Paolo Tenti is professor of Electronics and Power Electronics at the Department of Information Engineering of the University of Padova, Italy. His main research interests are in power electronics and industrial electronics. Currently, his work deals with power quality, distributed compensation techniques, and smart micro-grids, with special emphasis to optimum control of distributed power sources. He coordinated National research programs funded by MIUR in 2003-2006, 2005-2008, 2007-2010, and currently serves as a member of National and International evaluation panels, including EU ERC and Eurostar programs. From 1991 to 2001 he was a member of the Executive Board of the IEEE Industry Applications Society, and in 1997 he served as IAS President. From 2001 to 2008 he was Director of the Department of Information Engineering of the University of Padova. From 2003 to 2008 he served as Chairman of the Council of Department Directors of the same University. In 1999 he was elevated to the IEEE Fellow grade, and in 2000 was awarded the IEEE Millennium Medal.
Abstract: The success of machine learning algorithms heavily depends on data representation. Even the most performing classification algorithm will fail when provided with a bad input representation of data. In this talk, the problem of data representation for machine learning will be introduced. In particular, we will focus on kernel machines (e.g. support vector machines for classification and regression), the theory behind this kind of algorithms and well founded methods to learn the implicit representation of kernels from data.
Bio: Fabio Aiolli received a PhD in Computer Science in 2004 from the University of Pisa. He was Post-doc at the Computer Science Dept., University of Pisa (ITALY), Visiting Scholar at the University of Illinois at Urbana-Champaign (IL), USA, and Post-doc at the Dept. of Pure and Applied Mathematics, University of Padova (ITALY). Since 2006 he is Assistant Professor at the Dept. of Mathematics, University of Padova (ITALY). His research activity is in the area of Machine Learning and Pattern Recognition. In particular, he has expertise in kernel methods for structured data, kernel and representation learning, hierarchical representations and deep learning, with applications to recommender systems, neuroscience and biology
Abstract: Bayesian Networks are a powerful tool to represent probabilistic relationships between random variables. In these lectures I will introduce the formalism, its representational power and limitations, discuss exact and approximate inference techniques and present solutions for both parameter and structure learning.
Bio: Andrea Passerini graduated Magna Cum Laude in Computer Engineering at the University of Florence in 2000 and received his Ph.D. at the same University in 2004. He is currently Associate Professor at the Department of Information Engineering and Computer Science (DISI) of the University of Trento. His main research interests are in machine learning techniques, especially learning in domains with complex structures, relations and constraints. He recently got interested in “constructive machine learning” problems, where the goal is learning to synthesize new entities with the desired characteristics. He co-authored over ninety refereed papers. His h-index is 20 (Google Scholar, April 2017).
Abstract: The lecture starts with an overview of current silicon power electronics devices and the properties and existing problems. The new WBG materials siliconcarbide (SiC) and Galliumnitride (GaN) and the power electronics devices from these materials will be introduced and their main properties are discussed. Then the lecture will give a wide overview of the different application fields e.g. industrial, automotive, aerospace and energy conversion. The lecture is as well on the special challenges on packaging for high temperature, high power density and high reliability.
Bio: Dr. Kai Kriegel received a degree in electrical engineering and a degree in business administration from the RWTH Aachen University. He was working at the Institute of Ferrous Metallurgy of the RWTH Aachen University and received the Dr.-Ing. degree in 2001. Since 2001 he has been working at Siemens Corporate Technology in Munich as project manager on different R&D projects in the field of power electronics and drive technology, especially for automotive and aerospace applications.
Abstract: Recent theoretical and technical progress in artificial neural networks has significantly expanded the range of tasks that can be solved by machine intelligence. In particular, the advent of powerful parallel computing architectures based on graphic processing units, coupled with the availability of “big data”, now allows to create and train large-scale, hierarchical neural networks known as deep neural networks (LeCun, Bengio, & Hinton, 2015). These powerful learning systems achieve impressive performance in many challenging machine learning tasks, such as visual object recognition, speech processing and natural language understanding. In this seminar, I will review the main theoretical foundations of artificial neural networks, discussing both supervised and unsupervised forms of deep learning, and sequential architectures based on recurrent networks. I will then provide examples and case studies related to a variety of cognitive tasks, as well as their applications on difficult optimization problems. During a practical, “hands-on session” students will be encouraged to train and analyze their own models on a simple pattern recognition problem.
Bio: Alberto Testolin received the M.Sc. degree in computer science and the Ph.D. degree in cognitive science from the University of Padua, in 2011 and 2015, respectively. He is currently a post-doctoral researcher with the University of Padua, focusing on computational modeling of cognitive processes. His main interests include deep learning, recurrent neural networks and probabilistic generative models, which are applied to investigate visual processing and attentional mechanisms. During his PhD, he developed a neurocomputational framework to simulate cognitive functions based on unsupervised deep learning. Besided his primary interest in cognitive science, he works with computer scientists to improve the computational aspects of learning models, for example by exploiting parallel computing architectures (GPUs) or by inventing novel learning algorithms. He also collaborates with electronic engineers to improve telecommunication technologies by optimizing wireless and underwater data transmission. He has been recently invited as instructor to the annual meeting of the Cognitive Science Society, where he discussed his unsupervised modeling approach on a special workshop dedicated to deep learning.