BrainChip Akida NSoC Uses Spiking Neural Network Architecture

Neuromorphic computing specialist BrainChip is bringing its Akida Neuromorphic System-on-Chip (NSoC) — built on spiking neural network (SNN) architecture — to market.

Designed for applications such as advanced driver assistance systems (ADAS), autonomous vehicles, drones and vision-guided robotics, SNNs aim to bridge the gap between neuroscience and machine learning.

Spiking neural networks replace the math-intensive convolutions and back-propagation training methods of traditional convolutional neural networks (CNNs) with biologically inspired neuron functions and feed-forward training methodologies.

The Akida (the Greek word for “spike”) NSoC uses a CMOS logic process, and is designed for use as a standalone embedded accelerator or as a co-processor.

Embedded vision platforms, which use SNNs, are developing into a key component for fully autonomous vehicles — they give an automobile a set of eyes in the form of multiple cameras and image sensors.

Since self-driving vehicles will have to process an astronomical load of sensor data, SNNs, once trained, are better equipped to process constantly changing streams of information.

BrainChip claims each NSoC has effectively 1.2 million neurons and 10 billion synapses. Additionally, ingrained within the Akida neuron model are training methodologies for supervised and unsupervised training.

The Akida NSoC is designed to allow on chip training and off-chip training in the Akida Development Environment, a machine learning framework for the creation, training, and testing of SNNs.

The Development Environment includes the Akida Execution Engine, which contains a software simulation of the Akida neuron, synapses, and the multiple supported training methodologies.

The framework leverages the Python scripting language and its associated tools and libraries, including Jupyter notebooks, NumPy and Matplotlib.

The Akida NSoC includes sensor interfaces for traditional pixel-based imaging, dynamic vision sensors (DVS), Lidar, audio and analog signals, and boasts high-speed data interfaces such as PCI-Express, USB and Ethernet.

“Artificial intelligence at the edge is going to be as significant and prolific as the microcontroller. With the Akida NSoC, BrainChip is forging that path and leading the way,” CEO Lou DiNardo noted in a statement. “We are collaborating with major global manufacturers in a multi-market strategy to drive early adoption of the Akida NSoC.”

In addition, the chip’s scalability allows users to network many Akida devices together to perform neural network training and inferencing, the company noted.

— Nathan Eddy is a filmmaker and freelance journalist based in Berlin. Follow him on Twitter.

Leave a comment

Your email address will not be published. Required fields are marked *