27 Sep, 2016 - BrainChip Holdings Limited (ASX: BRN), developer of a revolutionaryembeddable semiconductor solution enablinghardware-based autonomous learning systems and applications, has signed a commercial marketing agreement with T2M,the world’s largest independent global semiconductor technology provider. Under the terms of the agreement T2M will act as a global representative of BrainChip’s technology solutions.
BrainChip’s ‘Spiking Neuron Adaptive Processor’ (SNAP) technology is a next generation hardware solution that acts like ‘digital neurons’, with the ability to learn autonomously, evolve and associate information just like the human brain. Peter van der Made, founder and CEO of BrainChip, comments, “Our SNAP technology is years ahead of deep learning; the latter is a simple accumulate and multiply software structure that has no biological equivalent whatsoever. Deep learning requires millions of samples and long ‘training’ periods that use an external optimizing technique on a fast computer. BrainChip SNAP on the other hand is fast and flexible, learning in seconds on the device itself.”
BrainChip’s SNAP learns by itself, without specific training, and finds patterns in the input stream that humans may not be aware of. This rapid learning capability opens up a whole new area of possibilities, to find images in a long video, patterns in stock market data, and thousands of other applications where deep learning is severely impaired.
The SNAP technology’s neurons autonomously learn through a process known as STDP (Synaptic Time Dependent Plasticity). The fully digital neurons process input spikes directly in hardware and are all updated in parallel, which means that the response time of the network is independent of the network size. Sensory neurons convert physical stimuli into spikes. Learning occurs when the input is intense, or repeating through feedback at each neuron and this is directly correlated to the way the brain learns.
Depending on power consumption requirements and network size, the SNAP technology can run faster than the speed of biological neurons and orders of magnitude faster than software neural networks.
The configurable technology is ideal for integration into portable devices for local processing of sensor data, enabling it to respond in real time with low latency, rather than having to send data to cloud-based computing.
An entire system can be embedded in a single chip, making it ideal for applications like pattern recognition or autonomous feature extraction. Typical applications where this could be embedded include smart phones, internet of things (IoT), robotics, driverless vehicles, drones and air transport, speech and image recognition and many other environments that can benefit from neural computations.
Since this is advanced state-of-the-art hardware, BrainChip chose T2M as its commercial marketing partner because of the company’s expertise in the global semiconductor industry and with delivering complex system-level solutions to manufacturers developing next generation mobile, IoT (internet of things), wearable technology products, consumer electronics, digital TV, security and image processing products.
“Thepartnership with T2M is another important step towards the integration of BrainChip’s IP into professional and consumer products worldwide,” said Peter van der Made. “T2M has extensive industry experience and contacts to help accelerate BrainChip’s route to market and continues our strategy of developing strong, worldwide partnerships with outstanding industry partners”.
T2M is the world’s largest independent global semiconductor technology provider, delivering complex system-level solutions for communications and consumer devices. Located in all the key tech clusters around the world, the company supplies state-of-the-art complex technology from tier one semiconductor providers for development of systems solution for mobile, IoT (internet of things), wearable technology products, consumer electronics, digital TV, security and image processing. It also delivers high-volume silicon-proven production IP as a complete SoC(system-on-chip) design database or individual IP blocks extracted from it. T2M was founded in 2010, and now has a team of 30 people located in key high-tech clusters in the USA, Europe, Europe, China, Taiwan, South Korea, Japan, Singapore and India. https://t-2-m.com/
About BrainChip Holdings Ltd (ASX:BRN)
BrainChipInc, located in Aliso Viejo, CA, has developed a revolutionary new Spiking Neuron Adaptive Processor (SNAP) hardware engine that has the ability to learn autonomously and rapidly, evolve and associate information just like the human brain. SNAPvision is a superior visual recognition toolkit and API. The technology is fast, completely digital, and consumes very low power. Additional information is available by visiting www.brainchipinc.com
About the SNAP technology
Spiking Neuron Adaptive Processor (‘SNAP’) is the next generation of machine learning that is embeddable.SNAP is delivered as IP, firmware or as a software emulation. SNAP learns within the device. The IP is unique in that it uses a processing core that is not programmed, but learns autonomously from sensory input streams, mimicking a human brain, and is considerably faster than deep learning’s convolutionalnetworks. It is able to learn in real time, which deep learning cannot do.
The SNAPhardware engine is delivered as configurable IP, a chip design that is embedded with a client’s own technology and other IP components into a chip that goes into a specific product. The clientpays for a license to use the SNAP technology and a royalty on each product that is produced – a business model that is also used by other highly successful companies around the world.
SNAP technology has been proved in several demonstrations, such as a simulated car that learned its own way around an unknown trackand the Autonomous Visual Feature Extractionsystem (AVFE) that was demonstrated with BrainChip’s Milestone 3 Client / Server access. The AVFE system proved that the technology learns from sensory input without training. ( http://tinyurl.com/hbe2g7x)
SNAPvision is a recent addition to SNAP, that enables SNAP to be applied insuperior visual recognition systems. With close to 100% accuracy, the instant learning of new features and a low power core, SNAPvision is ideal for use in cameras, robotic vision, and all types of surveillance systems. An application of SNAPvision can be seen here: http://www.brainchipinc.com/applications/videos/video/1336/people-track-loitering