An overview of the TrueNorth neuromorphic chip, software and applications

The Faculty of Informatics is pleased to announce a seminar given by Myron Flickner

DATE: Thursday, April 23rd, 2015
PLACE: IDSIA-SUPSI Manno, Anfiteatro (Galleria 2)
TIME: 11:00

ABSTRACT:
Brain inspired architectures offer tremendous promise for real-time low power operation. I will describe TrueNorth, a real-time neurosynaptic processor that implements a non-von Neumann, low-power, scalable, architecture. With 4096 neurosynaptic cores, the TrueNorth chip contains 1 million digital neurons and 256 million synapses tightly interconnected by an event-driven routing infrastructure.  The architecture is well suited to applications that use complex neural networks.  I will also describe the machine learning algorithms and software tooling we created to program the chip.  Finally some applications of the chip running video object recognition algorithms in real-time, while consuming only 63mW of power will be described.

BIO:
Myron "Flick" Flickner is an engineer, manager, inventor, programmer and author with over 20 patents and 75+ publications in areas of neuromorphic computing, image analysis, computer virus detection, retail,  and human-computer interaction.  Flick currently works at IBM Research as the manager of the software team creating brain inspired low power neuromorphic computer systems.  In 2010 and 2011 Flick was a Product Manager at Google.   From 2003 to 2010 Flick was the Global Retail Industry Leader for IBM Research.  He created and  managed Research projects relevant to the Retail industry and educated industry leaders on disruptive technologies.   A 31+ year IBMer,  Flick joined IBM Research San Jose in 1982 working on automated inspection of thin film disk heads. During 1987-1988 at IBM's Thomas J. Watson Research Center he researched image processing languages and served as IBM's representative to the ANSI X3H3.8/ISO (image processing) standards committee.  He returned to the Almaden Research Center and explored content based image retrieval and created the widely referenced QBIC (Query by Image Content) system.   After that he founded the BlueEyes project to explore how computing experiences change when the computer understand the emotional state of users.  This migrated into a project on Attentive Environments - sensor augmented spaces that are user and context aware - which became the early foundation for IBM's Smart Surveillance System (S3) offering.  His current research interests include software tools and algorithms for spiking neuromorphic systems.  He  received a B.S.(1980) and a M.S.(1982) in Electrical Engineering from Kansas State University.

HOST: Prof. Juergen Schmidhuber