Skip to main content

Neuromorphic computing and challenges

By
Category
Level
Intermediate

This workshop covers recent developments in this rapidly developing field of neuromorphic computing systems, and challenges ahead.

Workshop at Neuroinformatics 2015 in Cairns, Australia.

Workshop abstract

Future computing systems will capitalize on our increased understanding of the brain through the use of similar architectures and computational principles. During this workshop, we bring together recent developments in this rapidly developing field of neuromorphic computing systems, and also discuss challenges ahead.

In the neuromorphic systems field, emulation of neural systems is done using the implementation of neural elements in silicon. Typically, parallel analog and/or digital VLSI circuits are used; and the stochastic behavior of event driven communication between simple devices resembling neurons embedded in massively parallel and recursive network architectures is exploited. Such hardware systems, whose design is inspired by the brain, have the potential to create a paradigm shift in terms of energy efficiency, fault tolerance, adaptability as well as information processing capabilities.

For example, neuromorphic systems may in the future be able to mimic the capabilities of adaptive pattern recognition and motor control capabilities found in the vertebrate brain. Also, already today neuromorphic systems allow emulation and simulations of computational neural models in real time or faster.

 

Course Features
Simulation software for spatial model neurons and their networks designed primarily for GPUs
Past and present neurocomputing approaches
Experiences of students
Introduction to neuromorphic engineering
Lessons of this Course
1
1
Duration:
21:15

Presentation of a simulation software for spatial model neurons and their networks designed primarily for GPUs.

2
2
Duration:
41:57

Presentation of past and present neurocomputing approaches and hybrid analog/digital circuits that directly emulate the properties of neurons and synapses.

3
3
Duration:
20:39

Presentation of the Brian neural simulator, where models are defined directly by their mathematical equations and code is automatically generated for each specific target.

4
4
Duration:
19:57

The lecture covers a brief introduction to neuromorphic engineering, some of the neuromorphic networks that the speaker has developed, and their potential applications, particularly in machine learning.