This workshop is a venue for exchange of ideas and prognoses, and is meant to introduce engineers in industry, and researchers to what Neuromorphic computing can do for them.
Neuromorphic computing is an interdisciplinary approach to building physical computation structures that are brain-inspired. This approach can enable the deployment of AI and traditional algorithms, by dramatically reducing task–computation time and system energy consumption. While modern AI implements Neural Network abstractions in software, Neuromorphic hardware mimics Biological Neural Networks. Furthermore, Neuromorphic computing is much more than simply being an efficient hardware implementation of modern AI; it is defined by three ideas used separately or in combination: (i) the use of distributed, dynamic and efficient Spiking Neural Networks (SNNs), or Compute-in-Memory networks, (ii) the use of advanced materials and devices in addition to those based on Silicon, and (iii) application to real-time processing tasks, such as video/audio signal processing, motion planning, or data streaming.
This technology is used both for lowering the energy consumed for running heavy algorithms on computers, and also for higher-throughput interfacing of computers with the physical world. Development of neuromorphic architectures, integrated circuits and systems is an imminent research field for lowering energy requirements of the heavy computations needed in AI such as Deep learning algorithms, promising at least a ten thousand-fold drop in energy usage, and leading to real-time AI capacity in power-constrained and edge applications
