Session Overview |
Wednesday, May 29 |
10:40 |
Dimensionality Reduction in Photonics Design - New Methods and Applications
* Yuri Grinberg, National Research Council of Canada, Canada Dan-Xia Xu, National Research Council of Canada, Canada Muhammad Al-Digeil, National Research Council of Canada, Canada Daniele Melati, Centre de Nanosciences et de Nanotechnologies, CNRS, Universite Paris-Saclay, France Robert Hunter, SUNLAB, Nexus for Quantum Technologies Institute, University of Ottawa, Canada Alexandre Walker, National Research Council of Canada, Canada Gavin Forcade, SUNLAB, Nexus for Quantum Technologies Institute, University of Ottawa, Canada Jacob Krich, SUNLAB, Nexus for Quantum Technologies Institute, University of Ottawa, Canada Karin Hinzer, SUNLAB, Nexus for Quantum Technologies Institute, University of Ottawa, Canada Md Mahadi Masnad, Department of Electrical and Computer Engineering, McGill University, Canada Odile Liboiron-Ladouceur, Department of Electrical and Computer Engineering, McGill University, Canada Pavel Cheben, National Research Council of Canada, Canada Jens Schmid, National Research Council of Canada, Canada Siegfried Janz, National Research Council of Canada, Canada Dimensionality reduction (DR) has been an integral part of exploratory data analysis and feature selection in a multitude of machine learning applications. In particular, it has shown to be useful in multi-parameter photonics design problems where the objective function landscape offers a range of optimized designs. This talk will cover several recent advances in the methodology as well new design applications. Progress on sampling from data-efficient non-linear DR techniques will be presented, along with a method to reduce the computational load of data collection incurred by high fidelity solvers. New design problems in integrated silicon photonics as well as multijunction photonic power converters will be presented as the study cases. |
11:05 |
Novel photonic schemes for AI, and AI for band discovery
* Marin Soljacic, MIT, United States of America We will discuss opportunities for photonic p-bit computing, as well as certain novel AI schemes for materials discovery. |
11:40 |
Machine learning methods for designing and modeling photonic systems
* Jonathan Fan, Stanford University, United States of America We will discuss computational algorithms based on deep neural networks that can accelerate the design and simulation of nanophotonic devices, using metasurfaces and metamaterials as a model system. We will discuss the use of generative networks to perform population-based optimization and elucidate how the neural network architecture can be tailored to effectively perform freeform optimization. We will also discuss how physics-augmented deep networks can be trained with a combination of data and physics constraints to serve as accurate surrogate electromagnetic solvers. A principal challenge involves configuring the algorithms in a manner that enables application to a wide range of problems, and we show how these concepts can generalize to the simulation and optimization of photonic devices involving a range of domain sizes, fitting parameters, and functions. Together, these algorithms can effectively search for the global optimum three to four orders of magnitude faster than with conventional methods. We anticipate that with proper co-design of the neural network architecture with the scientific computing task, our surrogate solver and optimizer concepts can be adapted to large scale, three-dimensional photonic systems. |