Interface-like memristive device pushes neuromorphic computing forward – News

Researchers at the Los Alamos Laboratory’s Center for Integrated Nano Technologies (CINT) recently published details of a neuromorphic computing device that mimics the behavior of a neural synapse. The result? Computing that achieved 94.72% accuracy in recognizing handwritten numbers.

The device is an interface-type memristor made of gold/niobium-doped strontium titanate, all in a Schottky structure (Au/Nb:STO). The analog resistance of the device can be controlled via the memristor interface, and with these materials, Schottky barrier parameters, such as polarity and voltage magnitude, can be changed to change the device’s conductance.

Memristors are a promising technology in neuromorphic computing

Image used courtesy of the Integrated Nanotechnology Center

Memristors are a promising technology in neuromorphic computing since they can be programmed and “remembered” even when turned off. This mimics “synaptic plasticity,” which is an important basis in memory and learning in the brain. It allows synapses to strengthen or weaken based on their activity and is controlled by neurotransmitter receptors on the synapse.

In addition to synaptic plasticity, the researchers’ prototype can also mimic other synaptic functions such as coupled pulse facilitation, short-term potentiation and depression, long-term potentiation and depression, and spike timing-dependent plasticity. The Los Alamos team hypothesizes that their new device could circumvent traditional von Neumann bottleneck challenges.

Solving the von Neumann bottleneck

The von Neumann Bottleneck describes a problem in classical computer architecture where processing and memory are separate. To send information to a computer’s central processing unit (CPU) or graphics processing unit (GPU), the data must be read from memory and then transferred via a data bus.

The bottleneck occurs during this data transfer. Researchers have made significant efforts over the years to minimize this bottleneck, using strategies such as prefetching, speculative execution, or caching. However, data speeds are still limited to some extent, which can be a challenge when datasets such as images or videos need to be transferred and processed.

This data transfer consumes a lot of energy; In a world where data centers are used for applications such as machine learning, energy consumption is also a growing concern for both cost and environmental impact.

Memristor devices like the one devised by CINT have the potential to perform both data processing and storage in the same physical device, which not only overcomes the data transfer bottleneck, but can also reduce power consumption.

Meaning of MNIST results

The Modified National Standards and Technology (MNSIT) dataset of handwritten digits is often used as a benchmark for machine learning and image classification performance. The dataset contains a collection of 28 x 28 grayscale handwritten numbers from 0 to 9.

Interface-type memristor device performance results

Interface-type memristor device performance results. Image used courtesy of the Integrated Nanotechnology Center

The CINT team used a Crossbar Simulator to build a three-layered neural network and trained it using back propagation in 25 epochs using the synaptic functions of long-term potentiation and long-term depression. The simulation achieved a prediction accuracy of 94.72%. The researchers say this is superior to other candidate memristor architectures, such as conductive filament-type memristors.

Today’s benchmarks are in the ~99.8% accuracy range. However, the CINT results of 94.72% are still remarkable, considering this device is in its early research stages.

#Interfacelike #memristive #device #pushes #neuromorphic #computing #News
Image Source :

Leave a Comment