CFP

NIPU 2018: Neuromorphic Hardware in Practice and Use

Website	http://neuroscience.sandia.gov/research/wcci2018.html
Submission link	https://easychair.org/conferences/?conf=nipu2018
Submission deadline	May 7, 2018

Topics: neural networks machine learning neuromorphic computing

Organizers

Craig M. Vineyard, PhD

Sandia National Laboratories*

cmviney@sandia.gov

William M. Severa, PhD

Sandia National Laboratories*

wmsever@sandia.gov

Kristofor D. Carlson, PhD

BrainChip Inc.

kcarlson@brainchipinc.com

Website

http://neuroscience.sandia.gov/research/wcci2018.html

Description of the workshop

Abstract

This workshop is designed to explore the current advances, challenges and best practices for working with and implementing algorithms on neuromorphic hardware. Despite growing availability of prominent biologically inspired architectures and corresponding interest, practical guidelines and results are scattered and disparate. This leads to wasted repeated effort and poor exposure of state-of-the-art results. We collect cutting edge results from a variety of application spaces providing both an up-to-date, in-depth discussion for domain experts as well as an accessible starting point for newcomers.

Goals & Objectives

This workshop strives to bring together algorithm and architecture researchers and help facilitate how challenges each face can be overcome for mutual benefit. In particular, by focusing on neuromorphic hardware practice and use, an emphasis on understanding the strengths and weaknesses of these emerging approaches can help to identify and convey the significance of research developments. This overarching goal is intended to be addressed by the following workshop objectives:

1) Explore implemented or otherwise real-world usage of neuromorphic hardware platforms

2) Help develop 'best practices' for developing neuromorphic-ready algorithms and software

3) Bridge the gap between hardware design and theoretical algorithms

4) Begin to establish formal benchmarks to understand the significance and impact of neuromorphic architectures

Relevance to IEEE WCCI

IJCNN covers a wide range of topics in the field of neural networks and neural computation. In recent years, these topics have expanded to include biologically inspired hardware implementation research as well. The rapidly evolving need for hardware accelerators to enable the algorithmic and theoretical advances being made by WCCI attendees and participants also necessitates an understanding of how the interplay of algorithms and architectures in the form of neuromorphic computation is advantageous. A better understanding of the algorithm and architecture interplay also allows for the development of meaningful benchmarking which can further highlight the significance of research advances.

We expect three primary groups to comprise the audience.

Neuromorphic Hardware Experts: Recently a new field of neuromorphic platforms has become available, either at prototype or release stages. These platforms originate from a variety of groups within industry, academia and government. Hardware designers and platform stakeholders have a keen interest in early adoption, algorithms and applications as well as comparative and case studies.
Spiking Neural Network Algorithm Designers: As is evidenced by the growing presence of spiking neural networks at conferences such as IJCNN, biologically inspired spiking neural networks offer new and expanding capabilities. However, these algorithms are rarely designed following particular hardware constraints, and this creates a challenge when implementing the algorithms in practice. By expounding on neuromorphic implementations of spiking networks, algorithm designers stand to expand both utility and practicality.
Low-Power and Embedded Application Spaces: Neuromorphic platforms offer compelling improvements in performance-per-Watt. However, these numbers are often vendor-supplied and rarely include difficulties involved with algorithm porting. This workshop will offer a true-to-life story of the process, benefits and pitfalls of using neural networks on these up-and-coming platforms.

Scope and Topics:

Neuromorphic hardware; benchmarks and comparisons; applications, software, and toolkits; algorithms; workflows and integration

List of committed program committee members

James B. Aimone, PhD – Sandia National Laboratories

Conrad D. James, PhD – Sandia National Laboratories

Garrett Kenyon, PhD – Los Alamos National Laboratory

Dhireesha Kudithipudi, PhD - Rochester Institute of Technology

Catherine Schuman, PhD – Oak Ridge National Laboratory

Workshop duration, format, activities, and schedule

This half day workshop will consist of a brief overview of the various approaches to neuromorphic computation and motivate the need for applied results and benchmarks to properly characterize the significance of such approaches. Following the introduction will be a series of talks interleaving invited keynote speakers and contributed talks.

WT6 - FRIDAY, July 13th, 2:10-4:10pm (1st part); 4:30-6:30pm (2nd part) – ROOM ARUBA I

Time	Speaker	Title/Topic
2:10 – 2:40	William Severa & Sam Green (Sandia National Labs)	Welcome and Opening Overview Talk
2:40 – 3:10	Jeff Krichmar (UC Irvine)	CARLsim4: An Open Source Library for Large Scale, Biologically Detailed Spiking Neural Network Simulation using Heterogeneous Clusters
3:10 – 3:40	Hirak Kashyap (UC Irvine)	A Recurrent Neural Network Based Model of Predictive Smooth Pursuit Eye Movement in Primates
3:40 – 4:10	Dermot Kerr (University of Ulster)	From retina understanding to neuromorphic vision
4:10 - 4:30		Break
4:30 - 5:00	Michael Smith (Sandia National Labs)	Bridging Theory to Application in Neuromorphic Computing: Mapping the Liquid State Machine on to the STPU
5:00 – 5:30	Nicholas Soures (Rochester Institute of Technology)	Spiking Neural Networks for Classification and Prediction
5:30 – 6:00	Lyes Khacef (University Cote d’Azur)	Neuromorphic hardware as a self-organizing computing system
6:00 – 6:30		Open Discussion & Concluding Thoughts