Photonic-Electronic Memristors for Neuromorphic Applications
Details are subject to change.
-
Organiser
Alexandros Emboras, ETH Zurich, Switzerland
-
Chair
Nadia Jimenez Olalla, ETH Zurich, Switzerland
-
Day & Time
21.09.2022, 15:45 – 18:00
-
Location
Room Kairo
-
Description
Today’s artificial intelligence (AI) performance has been significantly improved thanks to the CMOS technology and the high computational power brought by graphics processing units (GPUs) and application specific integrated circuits (ASICs). However, to keep up with this trend, a critical problem should be solved, the inherent high energy consumption induced by the continuous exchange of data between the memory and computing units, which are physically separated. This issue is known as the “von Neuman bottleneck”.
Several innovations in the field of information technology have shown promise in overcoming this fundamental limit. For example, recent developments of memristors, a class of two-terminal nano-devices with a variable resistance, enables the collocation of the computing and storing functionalities, thus circumventing the limitations of current von Neumann designs. On the other hand, progress in standard photonic circuits allows for high-bandwidth optical data communication. Ideally, a photonic-electronic platform is desired that can simultaneously take advantage of the high density and non-volatility of electronic memristors and of the high-speed communication capabilities provided by photonics/plasmonics components. In this symposium, we will discuss the challenges and opportunities of this platform.
The symposium is divided in three sessions: Session 1 will cover the theoretical aspects related to the understanding of the interplay between photonic, electronic, phononic and ionic interactions within memristors. Session 2 will focus on the materials needed for novel memristive material stacks. Session 3 will be related to the device engineering and novel opto-electronic applications
-
Programme
15:45: Introduction
15:50: Materials, Thicknesses and Capping Layer Selection for Improved Memristive Properties
Ilia Valov, Research Centre Jülich, Germany16:10: Exploiting the Dynamics of Memristive Devices Based on the Valence Change Mechanism for Analog Computing
Stephan Menzel, Research Centre Jülich, Germany16:30: Outperforming Machine Learning, Through Biological Models with Memristive Analogues
Timoleon Moraitis, Huawei Technologies – Zurich Research Center, SwitzerlandBreak (10 min)
17:00: Closing the Gap Between Devices, Circuits, and Algorithms Towards Brain-inspired Edge Computing
Melika Payvand, University of Zurich, Switzerland & ETH Zurich, Switzerland17:20: A BaTiO3 ferroelectric multilevel non-volatile photonic phase shifter
Jacqueline Geler-Kremer, IBM Research−Europe, Switzerland17:40: Picosecond Time-Scale Resistive Switching Monitored in Real-Time
Miklos Csontos, ETH Zurich, Switzerland -
Supported by