This page is created to inform any interested participants regarding the seminars to be conducted for the EE690 Graduate Seminars course. All the seminars can be followed online via this Zoom link. The schedule and the content will be updated throughout the semester.

2023-2024 Fall Graduate Seminars

Seminar Schedule
Date Speaker Institution & Country Topic
Oct 17 Ekmel Özbay Bilkent University, Ankara, Türkiye How to convert the basic R&D activities in universities to 
technological commercial products through spin-off companies?
Oct 24 Sema Dumanlı Boğaziçi University, İstanbul, Türkiye In-vivo Sensing through Genetically Modified Bacteria
Nov 7 Ergin Atalar Bilkent University, Ankara, Türkiye Gradient Arrays for Magnetic Resonance Imaging
Nov 14 Figen S. Öktem Middle East Technical University, Ankara, Türkiye Computational Imaging and Physics-Informed Deep Learning: Making the 
Invisible Visible
Nov 21 Şahin K. Özdemir The Pennsylvania State University, University Park, PA, USA Non-Hermiticity as a Resource for Controlling Light and Its 
Interaction with Matter
Nov 28 Haluk Külah Middle East Technical University, Ankara, Türkiye Microsystems for Biomedical Applications
Dec 5 Cengiz Beşikçi Middle East Technical University, Ankara, Türkiye Quantum Structure Infrared Photon Sensors for New Generation Thermal Imagers
Dec 12 Arman Ayan EPFL, Lausanne, Switzerland All you need is light (and a nonlinear platform)!
Dec 26 İlkay Ulusoy Middle East Technical University, Ankara, Türkiye Deep Learning Models in the Presence of Noisy Labels

Title: How to convert the basic R&D activities in universities to technological commercial products through spin-off companies?

Abstract: Turning research and development (R&D) activities into viable commercial technologies has been a major goal in most of the academic institutions and universities around the world. This is quite a challenging task, as it requires all major players (i.e. university, companies and funding agencies) to work together in a coherent and planned manner.

In this talk, we will share our experience how we were able to achieve this goal. Based around the university-based R&D activities, we were able to develop multiple commercial technologies that are now widely used in various applications. These include GaN based high power RF transistors and Millimeter wave Integrated Circuits (MMICs), silicon photonics devices, distributed fiber-optic based sensors and navigation technologies.

Bio: Prof. Dr. Ekmel Ozbay received his B.S. degree in electrical engineering from the Middle East Technical University, Ankara, Turkey in 1987. He received M.S. and Ph. D. degrees from Stanford University in electrical engineering, in 1989 and 1992. He worked as a postdoc in Stanford University and later as a scientist in Iowa State University. He joined Bilkent University (Ankara, Turkey) in 1995, where he is currently a full professor in Physics and EEE Departments. In 2003, he founded Bilkent University Nanotechnology Research Center (NANOTAM) where he leads a R&D group working on nanophotonics, metamaterials, GaN based high power RF transistor and MMICs, distributed fiber optic  sensing systems, fiber optic based navigational sensors, and silicon photonics devices. He holds 100+ patents on various topics related to these technologies. He has published 630+ articles in SCI journals. His papers have received 21,0000+ SCI citations with an h-index of 67. He is the 1997 recipient of the Adolph Lomb Medal of  OSA and 2005 European Union Descartes Science award. He worked as an editor for Optics Letters, PNFA, SPIE JNP and IEEE JQE journals. He is a full member of Turkish Academy of Sciences (2001) and European Academy of Sciences (2005). He received the “Lifetime Achievement Award” from IEEE in 2020.  He is also the CEO of a spin-off company: AB-MicroNano Inc., which is founded to commercialize the technologies developed in NANOTAM.

Title: In-vivo Sensing through Genetically Modified Bacteria

Abstract: This talk presents a wireless in-vivo sensing system. The sensing system consists of a bio-hybrid implant and a wearable reader antenna pair. The bio-hybrid implant has two parts: a biodegradable implant antenna and genetically modified bacteria. The biodegradable implant antenna operates as a passive reflector and genetically modified bacteria control the degradation speed according to the presence of a specific molecule of interest. As the implant antenna degrades, changes in its geometry shift its resonant frequency. This shift is tracked by the wearable reader antennas. Therefore, the presence of the molecule of interest can be wirelessly tracked in vivo in real-time from outside the body.

Bio: Dr. Dumanlı received the B.Sc. degree in electrical and electronics engineering from Orta Doğu Teknik Üniversitesi, Ankara, Turkey, in 2006, and the Ph.D. degree from the University of Bristol, Bristol, U.K., in 2010. She was with Toshiba Research Europe, Bristol, as a Research Engineer and a Senior Research Engineer from 2010 to 2017.  She is currently an Associate Professor at Boğaziçi University, Istanbul, Turkey. She is the founder of the Boğaziçi University Antennas and Propagation Research Laboratory (BOUNtenna). She has a grant portfolio of more than €800K. She is the current chair of IEEE AP/MTT/EMC/ED Turkey Joint Chapter and URSI-TR Commission K and a board member of URSI Turkey. She is the recipient of the IEEE Antennas and Propagation Society 2022 Donald G. Dudley Jr. Undergraduate Teaching Award and three times recipient of the Bogazici University Excellence in Teaching Award. Her current research interests include antenna design for implantable and wearable devices, in-body sensing, bio-hybrid implant sensors, and multiscale communications.

Title: Gradient Arrays for Magnetic Resonance Imaging

Abstract: The gradient systems are the engines of magnetic resonance imaging (MRI) scanners. The conventional MRI scanners are equipped with three coil-amplifier pairs. Expensive and very powerful amplifiers drive the x-, y- and z-gradient coils. The switching speed and strength of the gradient field determine the image quality and speed, but the nerve stimulation limits these due to the induced electric field in the body. The distribution of the magnetic field and the body's position determine the electric field. In conventional systems, linear magnetic fields are generated in a large diameter of a spherical volume (DSV) independent of how patients are located inside the scanner. However, linear fields are not always necessary in a large DSV during imaging. The changing need for linearity cannot be obtained with conventional design. We propose to design a gradient array system that dynamically maximizes the nerve stimulation threshold by modulating the body's electric field distribution while providing the needed magnetic field distribution during imaging. The proposed design will also enable imaging in the conventional imaging mode with the large DSV to ensure compatibility with the existing technologies. Unlike the conventional systems, many inexpensive digitally controlled amplifiers will be in the close vicinity of the coil elements. Complete digital control of the system will enable maximum utilization of the available hardware, provide high-fidelity gradient fields, and generate the highest-quality brain images. Currently, MRI is the most powerful diagnostic imaging instrument. The wealth of diagnostic information is limited by available imaging time and cost. The gradient hardware limitations hold many exciting imaging techniques for accurately diagnosing devastating diseases. The proposed gradient array system will eliminate one of the critical barriers to further developing MRI technology.

Bio: Ergin Atalar is a Professor of Electrical and Electronics Engineering at Bilkent University. He is a graduate of the same university. Dr. Atalar spent almost half of his academic career at Johns Hopkins University. While he was at Hopkins, he became known for his contributions to Cardiovascular Magnetic Resonance Imaging and Interventional MRI. External grants supported his work. He became PIs of several NIH R01 grants. His patented inventions resulted in the formation of ClearPoint Neuro, Inc. (formerly known as Surgi-Vision Inc.). The company's products enabled over 6,000 MRI-guided brain surgeries as of 2023. He left Hopkins in 2005 as a Radiology, Biomedical Engineering, and Electrical and Computer Engineering professor. After returning to Turkiye in 2005, he switched his research focus to magnetic resonance engineering and its safety. He received three major grants to form UMRAM, the National Magnetic Resonance Research Center. He authored 126 peer-reviewed journal papers. His h-index is 70, according to Google Scholar, with more than 16400 citations (h-index=46 and 6500+ citations according to Scopus) as of October  2023. He holds 54 US patents and is fellows of the International Society of Magnetic Resonance in Medicine and the National Academy of Inventors, USA. He received the Science Award from the Scientific and Technological Research Council of Turkiye. He is members of the Academia Europaea and Science Academy, Turkiye.

Title: Computational Imaging and Physics-Informed Deep Learning: Making the Invisible Visible

Abstract: Computational imaging is a rapidly evolving interdisciplinary field awarded of many Nobel prizes. In computational imaging, digital  processing is employed in conjunction with a physical system to form images. That is, images are computationally formed from some indirect  measurements by solving an inverse problem. Driven by advances in signal processing techniques and faster computing platforms, this approach continuously yields the development of next-generation imaging systems in consumer electronics, defense industry, space physics, bioimaging and medicine. These imaging systems enable new forms of visual information, new imaging functionalities, and reduced hardware complexity and cost, that would be difficult, if not impossible, to achieve using traditional imaging. In this talk, first the fundamentals of computational imaging will be introduced. Then the recent role of deep learning in advancing computational imaging will be discussed with a focus on image reconstruction. With the recent advancements in machine learning, deep neural networks can be exploited to successfully solve a wide variety of inverse problems in  computational imaging. We will present physics-aware deep learning techniques developed in our group which combine the neural networks with analytical methods. Their performance will be illustrated through both simulated and experimental data for a sample application in three-dimensional microwave imaging. The results demonstrate that developed data-driven reconstruction techniques simultaneously achieve state-of-the-art image quality and low computational cost, and outperform their conventional counterparts by a large margin.

Bio: Dr. Oktem is an Associate Professor in the Department of Electrical and Electronics Engineering at Middle East Technical University (METU)  and the principal investigator of the Signal Processing And Computational sEnsing (SPACE) Lab. She received the B.S. and M.S. degrees in electrical and electronics engineering from Bilkent University, Ankara, Turkey, in 2007 and 2009, respectively, and the Ph.D. degree in electrical and computer engineering from the University of Illinois at Urbana-Champaign (UIUC), IL, USA, in 2014. At UIUC, she was selected to the “List of Teachers Ranked as Excellent by Their Students”, and was a recipient of NASA Earth and Space Science Fellowship and Professor Kung Chie Yeh Endowed Fellowship. Before joining METU, she was a Post-Doctoral Research Associate at the NASA Goddard Space Flight Center, where she worked on the development of novel spectral imaging techniques for high-resolution solar imaging. Her research spans the areas of computational imaging, inverse problems, statistical signal and image processing, machine learning, compressed sensing, and optical information processing. She  currently serves in IEEE Women in Signal Processing Sub-Committee on Grade Elevation, Nominations and Awards, and as Treasurer for the IEEE SPS Turkey Chapter. She is a member of IEEE, EURASIP, and Optica.

Title: Non-Hermiticity as a Resource for Controlling Light and Its Interaction with Matter

Abstract: Open physical systems are in continuous exchange of energy, information, and matter with their surroundings. A unique hallmark of open systems is their discrete spectral degeneracies known as exceptional points (EPs) where both the complex eigenfrequencies and associated eigenvectors coalesce. Recent years have seen tremendous progress in the theory and experimental implementations of EPs and associated concepts.  This progress has led to a host of intriguing new results in optics with promising potential applications in controlling the propagation of light and its interaction with matter. In this talk, we will briefly discuss the outcomes of research activities in our group within the framework of non-Hermitian photonics. We will first discuss the emergence of EPs and exceptional surfaces (ESs) - a hypersurface on which every point is an EP- in photonic systems and in the interaction of light with matter, and then use them as resources for chiral perfect absorption [1], optical sensing [2], and for topological control of intensity and phase of THz fields [3]. We will end the talk with a discussion of the opportunities and challenges in non-Hermitian photonics and how the concepts can branch out to other disciplines of science and technology [4].

[1] Soleymani et al. Nat Commun 13, 599 (2022)
[2] Chen et al. Nature 548, 192–196 (2017); Peng et al. (unpublished)
[3] Ergoktas et al. Science 376, 184 (2022).
[4] Zhang et al. Nature Communications 13, 6225 (2022)

Bio: Sahin K. Ozdemir is a professor of Engineering Science and Mechanics (ESM) at Pennsylvania State University. Before joining Penn State in 2017, he was with the Electrical and Systems Engineering Dept. at Washington University in St. Louis (2009-2018). Ozdemir received his B.Sc. (1992) and M.Sc. (1995) degrees from Middle East Technical University (METU), Turkey, and his Ph.D. (2000) degree from Shizuoka University, Japan, all in Electrical and Electronics Engineering. He worked with Japan Science and Technology Agency (JST) on various quantum information projects as a researcher (2000-2003), group leader (2004-2008) and senior scientist (2008-2009) and served as Specially Appointed Prof. at the Graduate University for Advanced Studies (2003-2004) and at Osaka University (2014-2019). Ozdemir has published 170+ articles in topics ranging from quantum optics, quantum information, and non-Hermitian physical systems to plasmonics, micro/nano-photonics, and sensors. He is named as “Highly Cited Researcher” by Web of Science (2019-2023). Ozdemir is a fellow of  Optica and IOP and a senior member of IEEE.

Title: Microsystems for Biomedical Applications

Abstract: Since the first introduction in 1970’s, MEMS technology is becoming popular in many different application areas, including military, automotive, and consumer electronics, as it provides cheap, small, and smart sensors and actuators. This technology is especially critical for biomedical applications, resulting in a new research area shortly called BioMEMS. Application areas of BioMEMS range from diagnostics to micro-fluidics, systems for drug delivery, tissue engineering, and  implantable systems.

There are various BioMEMS projects completed or currently on-going at ODTÜ, including microfluidic systems for liquid biopsy applications, organ-on-a-chip systems, MEMS based cochlear implants, dielectrophoresis chips for cell separation, gravimetric sensors for cancer cell detection, microvalves and pumps for lab-on-a-chip systems, and electrochemical sensors for bacteria detection. This presentation will introduce the research projects at ODTÜ on BioMEMS and biomedical microsystems.

Bio: Prof. Dr. Haluk Külah graduated from the Department of Electrical and Electronics Engineering at METU in 1996. He completed his master's degree in the same department in 1998. In 2003, he completed his Ph.D. in the Department of Electrical Engineering and Computer Science at the University of Michigan. Subsequently, he served as a postdoctoral researcher at the same university for a period. In 2004, Dr. Külah returned to the Department of Electrical and Electronics Engineering at METU as a faculty member, where he continues his work to this day.

Dr. Külah's research areas include Micro-Electro-Mechanical Systems (MEMS), micro-scale energy harvesters, microsystems for biomedical applications (BioMEMS), and analog and digital electronic circuit design for MEMS sensors. He has published over 220 articles and papers in international journals and conferences, holds 25 national/international patents, and has 5 international patent applications.

Dr. Külah has received several awards, including the 2009 Research Incentive Award from the METU Prof. Dr. Mustafa Parlar Education and Research Foundation, the 2013 TÜBİTAK Incentive Award, the 2013 IBM Faculty Award, the 2015 BAGEP Young Scientist Award, and the 2020 Elginkan Foundation Technology Award. In 2015, he won the ERC Consolidator Grant for the FLAMENCO project, focusing on next-generation cochlear implants. He continued this project with the OPERA in 2020 and ARIA in 2023, receiving two separate ERC Proof of Concept Grants. Dr. Külah is a member of the Science Academy Turkey and also serves as a board member of the ODTÜ MEMS Center.

Title: Quantum Structure Infrared Photon Sensors for New Generation Thermal Imagers

Abstract: Utilization of band gap engineering with mature III-V semiconductor technologies has led to the development of novel infrared photodetectors based on quantum structured semiconductor systems as an alternative to conventional sensors fabricated with low energy band gap semiconductor materials. The Quantum Structure Infrared Photodetector (QSIP) family has reached the level of highly sensitive very large format infrared focal plane arrays and thermal imaging systems that are now commercially available.  This talk will cover the recent developments in quantum structured infrared photodetector  technology, as well as the studies at the Middle East Technical University toward the realization of larger format/lower cost single- and dual-band infrared sensors for new generation thermal imagers.

Bio: Cengiz Beşikçi received the Ph.D. degree in electrical engineering from Northwestern University in 1994. He is currently a Professor with Electrical and Electronics Engineering Department, Middle East Technical University where he founded the Quantum Devices and Nanophotonics Research Laboratory known to be one of the leading research laboratories on infrared sensors for thermal imaging. He is a well known scientist in the field of infrared imaging sensors, and he has published many scientific papers which  have significantly contributed to the development of infrared photon sensor technologies. He directed 14 large-scale research projects at METU, and he has been the prominent researcher in the development and establishment of related technologies and production facilities by the industry in Turkey. He was a recipient of numerous awards including Research Incentive Awards by TÜBİTAK and Prof. Dr. Mustafa Parlar Research and Education Foundation, four METU Thesis of the Year Awards (as thesis advisor), Cabell Fellowship by Northwestern University and various academic achievement awards. He chaired the prestigious Quantum Structure Infrared Photodetector (QSIP) International Conference in 2010. He is the founder and CEO of Optonik Semiconductor Technologies LLC.

Title: All you need is light (and a nonlinear platform)!

Abstract: In the presence of strong electric fields, the medium's dipole response shifts from linear to nonlinear, enabling control of light based on power, phase, or polarization. Nonlinear optics promises revolutionary capabilities for the transition from electronics to all-optical signal processing, fueling the development of efficient system-on-a-chip (SoC) designs. Recent breakthroughs in this field have also broadened applications in molecular sensing and spectroscopy. Diverse materials harness on-chip nonlinear optical phenomena, each presenting unique advantages and challenges. This presentation explores different aspects of nonlinear waveguide photonics, focusing on various nonlinear platforms that are competing for on-chip implementation of diverse applications leveraging distinct nonlinear interactions. While introducing a variety of nonlinear optical phenomena, the primary focus remains on exploring the potential of four-wave mixing. How waveguide engineering plays a role in spectral bandwidth will be addressed in pursuit of efficient broadband wavelength conversion and coherent light generation beyond the reach of common light sources.

Bio:  Arman Ayan earned his B.Sc. degrees in Electrical and Electronics Engineering and Physics from METU in 2015 and 2016, respectively. Subsequently, he completed his M.Sc. in Electrical and Electronics Engineering at the same university in 2018, dedicating over three years as a research assistant. His master's research primarily focused on the epitaxial growth of germanium on silicon and germanium-based photonic devices. For the past four and a half years, he has been a doctoral assistant at École Polytechnique Fédérale de Lausanne (EPFL), focusing on the nonlinear characterization of chalcogenide  photonic-crystal fibers, polarization-insensitive four-wave-mixing, and efficient-broadband four-wave-mixing in silicon nitride  waveguides. He successfully defended his doctoral thesis in November 2023.

Title: Deep Learning Models in the Presence of Noisy Labels

Abstract: Image classification systems recently made a giant leap with the advancement of deep neural networks. However, these systems require an excessive amount of labeled data to be adequately trained. Gathering a correctly annotated dataset is not always feasible due to practical challenges, especially for medical applications and label noise is a common problem in real-world datasets and applications. Although deep neural networks are known to be relatively robust to label noise, their tendency to overfit data makes them vulnerable to memorizing even random noise. Therefore, it is crucial to consider the existence of label noise and develop counter algorithms to fade away its adverse effects to train deep neural networks efficiently. In this talk, the problem of noisy label will be explained, possible solution approaches will be summarized and a novel label noise robust learning algorithm, MetaLabelNet, will be presented. The algorithm generates soft labels for each instance according to a meta-objective, which is to minimize the loss on the small meta-data. Afterward, the  main classifier is trained on these generated soft-labels instead of given noisy labels. In each iteration, before conventional learning,  the proposed meta objective reshapes the loss function so that resulting gradient updates would lead to model parameters with the minimum loss on meta-data.

Bio:  Ilkay Ulusoy was born in Ankara, Turkey, in 1972. She received the B.Sc. degree from the Electrical and Electronics Engineering Department, Middle East Technical University (METU), Ankara, in 1994, the M.Sc. degree from The Ohio State University, Columbus, OH, USA, in 1996, and the Ph.D. degree from METU in 2003. She did her research at the Computer Science Department, University of York, York, U.K., and Microsoft Research Cambridge, U.K. She has been a Faculty Member with the Department of Electrical and Electronics Engineering, METU, since 2003 and the department chair since 2019. Her main research interests are computer vision, machine learning, probabilistic graphical models and application of these to neuroscience.


Son Güncelleme:
22/12/2023 - 11:13