Neuromorphic Computing Special Sessions

NUO XU; Samsung (Special Sessions Organizer)

As Dennard scaling is almost approaching its limits, a new computing paradigm involving system, circuit and technology co-innovation is on demand. The “Brain-like” computation, a.k.a. the neuromorphic and neuro-inspired computing concepts have been considered as the most promising candidate as a replacement or supplement to existing Van Neumann computer architectures. Both emerging semiconductor devices and circuit/system designs have been studied extensively and many progress has been achieved in recent years. This two special sessions invite leading researchers in this field yet with different paths to implement a “Silicon” neural network

Session 17 – Neuromorphic Computing I: Device and Integration Technology
Philip H.-S. Wong; Stanford University
Shimeng Yu; Arizona State University
Arvind Kumar; IBM Research, Yorktown Heights
Zhe Wan; University of California at Los Angeles

In Session 17, Prof. Wong and Prof. Yu will introduce their work on emerging non-volatile memory based cognitive computing infrastructures, such as Resistive RAM (RRAM) and Phase Change Memory (PCM) based synapses. These technologies are extremely promising for energy-efficient computing due to their superior energy-delay products. The other two talks given by Dr. Kumar and Dr. Wan are aiming at 3-D wafer-scale integration technology for a high density implementation of semiconductor devices, in order to emulate networks with large connectivity, i.e. at a size of biological neural system.

Session 19 – Neuromorphic Computing II: Circuits and Systems
Giacomo Indiveri; ETH Zurich & University of Zurich
John Arthur; IBM Research, Almaden
Zhengya Zhang; University of Michigan at Ann Arbor

In Session 19, on the other hand, Prof. Indiveri will talk about the analog/mixed-signal circuit designs to implement neurons for different computational models. Dr. Arthur will provide a review of the IBM’s TrueNorth development work to demonstrate the feasibility of spiking neural network-based computing for commercial purposes. Prof. Zhang’s team has also designed spiking-based recurrent neural network as accelerators for computer vision applications; he will present this work to show the energy and performance merits of this new and “intelligent” hardware architecture

(c) Copyright 2015 Joyce Lloyd, IMF