Nordic Edge AI Lab is the gateway to ultra-efficient edge intelligence—an end-to-end automated development platform designed specifically for engineers. It enables developers to build efficient Neuton models and deploy them directly as value-added features on Nordic’s wireless SoCs and SiPs, revolutionizing the development paradigm for edge AI.
Unlike traditional AI model-building methods, Nordic Edge AI Lab employs a radically different, patented technology:
• Neuron-by-Neuron Construction: Rather than starting with a predefined large structure and then compressing it, the platform builds the neural network from scratch, automatically determining the optimal architecture during the learning process, one neuron at a time.
• Elimination of Redundancy: Combined with a patented global optimization algorithm, the platform circumvents the limitations of traditional gradient descent methods to generate unique networks free of redundant neurons and connections.
• Extreme Lightweight Design: Models created using this framework have an average memory footprint of less than 5 KB—10 times smaller than TensorFlow Lite models—achieving “extreme compactness” without sacrificing accuracy.
Dual Acceleration Technology Combining Software and Hardware
This solution not only provides model algorithms but also addresses the wide range of device requirements in industrial manufacturing through two complementary technologies:
• Neuton Models (Universal Adaptability): Optimized ultra-compact models are perfectly adapted to any Nordic SoC, enabling complex edge AI tasks to run on standard CPU cores.
• Axon NPU (Dedicated Acceleration): For more complex AI challenges—such as high-frequency time-series data or image classification—the solution deeply integrates dedicated AI accelerator cores, delivering inference speeds 15 times faster than running on a CPU.
Fully Automated, No-Code Development
This platform significantly lowers the barrier to entry for AI. Users do not need an in-depth background in data science; they simply need to upload high-quality sensor datasets, and the platform will automatically complete feature engineering and model training within minutes, truly enabling agile development with a “data-in, model-out” approach.
Malicious vote manipulation is expressly forbidden in this voting event. The organizers reserve the right to evaluate the fairness and accuracy of the voting results. AspenCore retains the authority to interpret the rules of this event.