A research group from Tohoku University has developed spintronics devices which are promising for future energy-efficient and adoptive computing systems, as they behave like neurons and synapses in the human brain.
Today’s information society is built on digital computers that have evolved drastically for half a century and are capable of executing complicated tasks reliably. The human brain, by contrast, operates under very limited power and is capable of executing complex tasks efficiently using an architecture that is vastly different from that of digital computers.
So the development of computing schemes or hardware inspired by the processing of information in the brain is of broad interest to scientists in fields ranging from physics, chemistry, material science and mathematics, to electronics and computer science.
In computing, there are various ways to implement the processing of information by a brain. Spiking neural network is a kind of implementation method which closely mimics the brain’s architecture and temporal information processing. Successful implementation of spiking neural network requires dedicated hardware with artificial neurons and synapses that are designed to exhibit the dynamics of biological neurons and synapses.
Here, the artificial neuron and synapse would ideally be made of the same material system and operated under the same working principle. However, this has been a challenging issue due to the fundamentally different nature of the neuron and synapse in biological neural networks.
The research group – which includes Professor Hideo Ohno (currently the university president), Associate Professor Shunsuke Fukami, Dr. Aleksandr Kurenkov and Professor Yoshihiko Horio – created an artificial neuron and synapse by using spintronics technology. Spintronics is an academic field that aims to simultaneously use an electron’s electric (charge) and magnetic (spin) properties.
The research group had previously developed a functional material system consisting of antiferromagnetic and ferromagnetic materials. This time, they prepared artificial neuronal and synaptic devices microfabricated from the material system, which demonstrated fundamental behavior of biological neuron and synapse – leaky integrate-and-fire and spike-timing-dependent plasticity, respectively – based on the same concept of spintronics.
The spiking neural network is known to be advantageous over today’s artificial intelligence for the processing and prediction of temporal information. Expansion of the developed technology to unit-circuit, block and system levels is expected to lead to computers that can process time-varying information such as voice and video with a small amount of power or edge devices that have the an ability to adopt users and the environment through usage.
The Latest on: Spiking neural network
via Google News
The Latest on: Spiking neural network
- Breakthrough in energy efficient artificial intelligenceon October 7, 2020 at 11:34 pm
Thanks to a mathematical breakthrough, AI applications can become a hundred to a thousand times more energy efficient. This will make it possible to put much more elaborate AI in chips, enabling ...
- Tenstorrent Is Changing the Way We Think About AI Chipson September 27, 2020 at 5:00 pm
On the flip side you've got the spiking artificial neural network, which is a lot less popular and has had a lot less success in in broad applications.” Spiking neural networks (SNNs) more closely ...
- Adaptive Cardiac Resynchronization Therapy Device: A Simulation Reporton September 20, 2020 at 5:01 pm
The adaptive CRT module is built from a microcontroller, pulse generator module, and a spiking neural network co-processor. The spiking neural network co-processor is the learning module that ...
- Dose-dependent effects of transcranial alternating current stimulation on spike timing in awake nonhuman primateson September 18, 2020 at 1:59 am
These authors contributed equally to this work. See allHide authors and affiliations Weak extracellular electric fields can influence spike timing in neural networks. Approaches to noninvasively ...
- Neural Networks Without Matrix Mathon September 16, 2020 at 5:00 pm
In this view, further advances require new learning models and new training algorithms. Spiking neural networks are one frequently discussed alternative, and spike-timing dependent plasticity is often ...
- A new strategy for the viral manipulation of interneurons in mice and other mammalson September 16, 2020 at 6:38 am
fast-spiking interneurons," Jordane Dimidschstein, head of the team who led the study, told Medical Xpress. "These neurons exert a strong level of inhibitory control over local networks ...
- Neural network IP Listingon September 7, 2020 at 5:00 pm
Inspired by the biological function of neurons but engineered on a digital logic process, this event-based spiking ... Efficient and Versatile Computer Vision, Image, Voice, Natural Language, Neural ...
- SHF:Small:Scalable Spiking Neural Network Enabled by Probabilistic and Non-Volatile Synapseson August 30, 2020 at 5:00 pm
Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge ...
- Tachyum Prodigy Native AI Supports TensorFlow and PyTorchon August 26, 2020 at 5:03 am
Tachyum's Prodigy can run HPC applications, convolution AI, explainable AI, general AI, bio AI and spiking neural networks, as well as normal data center workloads on a single homogeneous ...
via Bing News