BrainChip, the Australian spiking neural network specialist, started shipping evaluation boards last month.

The board complements the Company’s Akida Development Environment (ADE) for Early Access Partners (EAP) that have developed Akida compatible networks.
The ADE is a development environment that allows potential customers to design a neural network as a Convolutional Neural Network (CNN) and utilize the ADE workflow to convert the network to an Event-Based CNN or develop a native Spiking Neural Network (SNN).
The effect of implementing an Event-Based CNN is to significantly reduce power consumption by processing only non-zero activations and take advantage of the sparsity in most data. Additional power savings are captured by aggressive quantization without a meaningful loss of accuracy for most network designs.

Once converted to an Event-Based CNN network, a designer can quantize both weights and activations to 1, 2 or 4 bits selectively.
The Akida neural fabric provides significant new features that allow true AI Edge Learning rather than simple AI Edge Training, which requires connectivity through the cloud or a data center to re-train a network as new classifiers are detected. Additionally, the ability to implement AI Edge Incremental Learning is unique to the Akida NSoC and allows potential customers to deploy a new and valuable feature in the AI Edge products.
“True AI Edge learning does not exist with current AI solutions and our ability to provide this and other features while significantly reducing both power consumption, size and the Bill-of-Materials has attracted the attention of leading suppliers in the Smart Home, Smart Transportation and Smart City markets,” says BrainChip CEO Lou di Nardo.