In today’s landscape of Artificial Intelligence, Deep Learning and its numerous inference engines are monopolizing the front stage, but other technologies have essential benefits such as field training and real-time adaptivity, novelty detection, learning causality and traceability.
Among them, the NeuroMem NM500 chip is a digital neural network chip capable of intrinsic learning and recognition of patterns derived from multimedia such as images and sounds, but also instruments, text and other data types. Manufactured by nepes (Korea) under a license from General Vision, the NM500 features 576 neurons which can be trained on small datasets and cleverly tuned to deliver the best compromise between throughput and accuracy for a given application. For example, one may prefer to train a NeuroMem network acknowledging when it is uncertain or even ignorant rather than guessing or reporting a “closest” match which can still be quite far. This is made possible when the neurons are used as a Radial Basis Function classifier, and not as the commonly known K-Nearest Neighbor. It is this notion of ignorance and uncertainty which can trigger the intelligent decision to learn more or to have the wise recourse to another opinion. By combining multiple NeuroMem networks (or experts) trained differently on the same subject, accurate decisions can be made taking advantage of their complementarity or their domains of mutual exclusivity.
To experiment with NeuroMem networks, General Vision’s NeuroMem Knowledge Builder is a simple framework to train and test the neurons on your datasets while producing rich diagnostics. The company also offers simple APIs and tools for generic pattern recognition as well as image recognition. They all integrate a cycle accurate simulation of 8000 neurons and can also interface to a hardware NeuroMem network such as the Brilliant USB dongle (2304 neurons), the Arduino-compatible NeuroShield board (576 neurons) with expandable network capacity, and soon a cognitive SSD with high speed throughput and high network capacity. In addition to the NM500 chip, NeuroMem is available as an IP core for FPGAs from Xilinx, Altera and Lattice and also for licensing.
Dr. Manan Suri, professor at the Indian Institute of Technology, Delhi and his team from the Department of Electrical Engineering have conceptualized and qualified a system combining two NeuroMem neural networks to accurately authenticate persons based on their voice and face. The hardware platform includes a NeuroMem CM1K chip so its 1024 neurons can perform the learning and classification of patterns in real-time and near sensor.
General Vision was invited to participate to a webcast organized by QuickLogic on a new AI Ecosystem targeting end points. Indeed, with the advent of Industry 4.0 and the proliferation of IoT, there is a need to make devices that process sensor data and make decisions locally. Traditional machine learning algorithms are too memory and power hungry to migrate AI to endpoints with constrained battery life. This calls for NeuroMem neurons with zero latency, low power, low cost and easy to train.
Among the participants were nepes, manufacturer of the NeuroMem NM500 chip, and SensiML, developer of smart sensor algorithms. During his speech, Guy Paillet, CEO of General Vision, emphasized the synergy between the NeuroMem trainable neurons and eFPGAs to design ultra low-power and adaptive search engines. QuickLogic introduced their Quick AI HDK platform integrating their EOS SOC and a network of 2 NM500 chips.
Oceanit teamed up with Kauai Coffee Company and Kamehameha Schools to develop an innovation that can ‘see’ the ripeness of coffee cherries utilizing a NeuroMem neural network and win Hawaii’s first annual AGathon. By using a portable prediction system to determine ripeness, Kauai Coffee’s harvest values could be improved by more than a quarter million dollars per harvest. Read the complete report.
The Oceanit team developed a system around the Raspberry PI equipped with RaspiCam vision module and a shield board populated with two NeuroMem NM500 chips.
Nepes AI, a business unit of nepes (South Korea), launches the NM500 chip, the smallest neuromorphic IC available on the market and featuring no less than a NeuroMem neural network with 576 neurons. The chip can be integrated into sensors and IOT platforms to deploy pattern learning and recognition at the edge. Conveniently packaged in a wafer scale package, the chip delivers 120,000 pattern recognitions per second for less than 125 mW. Early adopters can put the neurons into use immediately with nepes’ NeuroShield module interfacing to the low-cost Arduino microcontroller boards through SPI.