5. A SoC With a Neural Network
The ST paper entitled A New Scalable Architecture to Accelerate Deep Convolutional Neural Networks for Low Power IoT Applications gained a lot of traction because it demonstrated that it is possible to run a neural network for image recognition applications within a SoC that integrates a Cortex-M4 and 128 KB of memory. Traditionally, a majority of people associate deep learning with powerful computers hosting multiple discrete and expensive graphics cards. The findings of our research teams were thus particularly groundbreaking because they dispelled this preconceived idea to open the door to a whole new computing paradigm.