T. Ferreira de Lima (US) Princeton University
Princeton UniversityAuthor Of 2 Presentations
MD2.4 - NEUROMORPHIC PHOTONICS FOR DEEP LEARNING
- V. Bangari (CA) Queen's University
- B. Marquez (CA) Queen's University
- A. Tait (US) National Institute of Standards and Technology
- M. Nahmias (US) Princeton University
- T. Ferreira de Lima (US) Princeton University
- H. Peng (US) Princeton University
- P. Prucnal (US) Princeton University
- B. Shastri (CA) Queen's University, Physics, Engineering Physics & Astronomy
Abstract
Abstract
Co-integrated neuromorphic photonic and electronic processors promise orders of magnitude improvements in both speed and energy efficiency over purely digital electronic approaches. We discuss neuromorphic photonic systems and their application to deep convolutional neural networks inference.
WP10 - AUTOMATED CONTROL OF THE TRANSFER FUNCTION OF AN INTEGRATED CASCADED MACH-ZEHNDER INTERFEROMETER
Abstract
Abstract
Thermal crosstalk and fabrication variations are detrimental to accurate control of integrated photonic systems. We build a calibration-based control algorithm accounting for these factors in a dual Mach-Zehnder Interferometer (MZI) structure, and experimentally demonstrate automated control of optical transmission with only 3% error.
Presenter Of 1 Presentation
MD2.4 - NEUROMORPHIC PHOTONICS FOR DEEP LEARNING
- V. Bangari (CA) Queen's University
- B. Marquez (CA) Queen's University
- A. Tait (US) National Institute of Standards and Technology
- M. Nahmias (US) Princeton University
- T. Ferreira de Lima (US) Princeton University
- H. Peng (US) Princeton University
- P. Prucnal (US) Princeton University
- B. Shastri (CA) Queen's University, Physics, Engineering Physics & Astronomy
Abstract
Abstract
Co-integrated neuromorphic photonic and electronic processors promise orders of magnitude improvements in both speed and energy efficiency over purely digital electronic approaches. We discuss neuromorphic photonic systems and their application to deep convolutional neural networks inference.