Time Domain Adjoint Sensitivities and their Applications: State of the Art
31 August, 2011
Speaker:Â Â Dr. Prof. Bakr from the Department of Electrical and Computer Engineering, McMaster University, Hamilton, Ontario, Canada
Date:Â Friday September 16, 2011
Time: Registration and Networking: 10:00 a.m..; Seminar: 10:30 a.m. – 11:30 a.m.
Location:Â University of Ottawa, School of Electrical Engineering and Computer Science, SITE Building, Room 5084 (Boarding Room), 800 King Edward Avenue, Ottawa, Ontario, Canada
Organizers:Â The IEEE Ottawa Communications Society, Broadcast Technology Society, and Consumer Electronics Society (ComSoc / BTS / CES) Joint Chapter, Antennas and Propagation Society and Microwave Theory & Techniques Society (AP / MTT) Joint Chapter, IEEE Photonics Society (PHO) Ottawa Chapter and IEEE Ottawa Section (OS).
ADMISSION: Free. Registration required. To ensure a seat, please register by e-mail contacting: Qingsheng Zeng or Wahab Almuhtadi.
Date:Â Friday September 16, 2011
Time: Registration and Networking: 10:00 a.m..; Seminar: 10:30 a.m. – 11:30 a.m.
Location:Â University of Ottawa, School of Electrical Engineering and Computer Science, SITE Building, Room 5084 (Boarding Room), 800 King Edward Avenue, Ottawa, Ontario, Canada
Organizers:Â The IEEE Ottawa Communications Society, Broadcast Technology Society, and Consumer Electronics Society (ComSoc / BTS / CES) Joint Chapter, Antennas and Propagation Society and Microwave Theory & Techniques Society (AP / MTT) Joint Chapter, IEEE Photonics Society (PHO) Ottawa Chapter and IEEE Ottawa Section (OS).
ADMISSION: Free. Registration required. To ensure a seat, please register by e-mail contacting: Qingsheng Zeng or Wahab Almuhtadi.
Abstract:
The design process of high frequency structures is usually carried out using Electromagnetic (EM) simulators.   A model of the structure under consideration is constructed and a number of key variables controlling its response are chosen.  An optimization algorithm (optimizer) drives the simulator to determine the optimal set of values of the designable parameters that satisfies the design specifications. Gradient-based optimizers are robust with well established convergence proofs.  They, however, require sensitivity information which may require large number of extra simulations for each design step.
The adjoint variable methods (AVM), aim at efficiently estimating the response sensitivities.  Using at most one extra EM simulation of an adjoint system, the response sensitivities with respect to all parameters are estimated regardless of the number of parameters.  For the case of network parameters, this extra simulation can be eliminated.  The same simulations supplying the network parameters supply their sensitivities as well.  This makes gradient-based optimization more efficient.I
n this talk we review the state of the art of the time-domain AVMs and their applications.  We discuss recent techniques that make this approach more efficient in terms of speed and memory storage.  We show a number of interesting applications in microwave imaging, antenna design, and design of photonic devices.  Open points for research are also addressed.
The design process of high frequency structures is usually carried out using Electromagnetic (EM) simulators.   A model of the structure under consideration is constructed and a number of key variables controlling its response are chosen.  An optimization algorithm (optimizer) drives the simulator to determine the optimal set of values of the designable parameters that satisfies the design specifications. Gradient-based optimizers are robust with well established convergence proofs.  They, however, require sensitivity information which may require large number of extra simulations for each design step.
The adjoint variable methods (AVM), aim at efficiently estimating the response sensitivities.  Using at most one extra EM simulation of an adjoint system, the response sensitivities with respect to all parameters are estimated regardless of the number of parameters.  For the case of network parameters, this extra simulation can be eliminated.  The same simulations supplying the network parameters supply their sensitivities as well.  This makes gradient-based optimization more efficient.I
n this talk we review the state of the art of the time-domain AVMs and their applications.  We discuss recent techniques that make this approach more efficient in terms of speed and memory storage.  We show a number of interesting applications in microwave imaging, antenna design, and design of photonic devices.  Open points for research are also addressed.
About the Speaker:
Mohamed H. Bakr received  a B.Sc. and M.Sc. degrees in Electronics and Communications Engineering and Engineering Mathematics from Cairo University, Egypt in 1992 and 1996, respectively with distinction (honors).  He earned the Ph.D. degree in September 2000 from the Department of Electrical and Computer Engineering, McMaster University.  In November 2000, he joined the Computational Electromagnetics Research Laboratory (CERL), University of Victoria, Victoria, Canada as an NSERC Post Doctoral Fellow.  Between July 2008 and June 2009, he was with Research In Motion (RIM) as a senior researcher during his Sabbatical leave.  His research areas of interest include computer-aided design and modeling of microwave and photonic circuits, neural network applications, efficient optimization using time/frequency domain methods, and bioelectromagnetism. He is a recipient of a Premier’s Research Excellence Award (PREA) from the province of Ontario in 2003, and a Discovery Accelerator Award (DAS) in 2011.  He is currently an associate professor with the Department of Electrical and Computer Engineering, McMaster University.
Mohamed H. Bakr received  a B.Sc. and M.Sc. degrees in Electronics and Communications Engineering and Engineering Mathematics from Cairo University, Egypt in 1992 and 1996, respectively with distinction (honors).  He earned the Ph.D. degree in September 2000 from the Department of Electrical and Computer Engineering, McMaster University.  In November 2000, he joined the Computational Electromagnetics Research Laboratory (CERL), University of Victoria, Victoria, Canada as an NSERC Post Doctoral Fellow.  Between July 2008 and June 2009, he was with Research In Motion (RIM) as a senior researcher during his Sabbatical leave.  His research areas of interest include computer-aided design and modeling of microwave and photonic circuits, neural network applications, efficient optimization using time/frequency domain methods, and bioelectromagnetism. He is a recipient of a Premier’s Research Excellence Award (PREA) from the province of Ontario in 2003, and a Discovery Accelerator Award (DAS) in 2011.  He is currently an associate professor with the Department of Electrical and Computer Engineering, McMaster University.