Co-Design Crosscut
The Co-Design crosscut thrust focuses on the development, testing, and refinement of a framework of methods to jointly co-optimize applications, algorithms, architectures, and technologies. Members of this thrust work together with members of other thrusts to enable the creation of this framework. In this thrust, we will work across the complete stack, from materials to algorithms, creating a complete co-design workflow.
At the materials level, we will create a hierarchy of simulations and AI and machine learning methods to predict and design properties of new electronic and photonic materials. At the devices level, we will develop cross-layer models for neuromorphic devices and use ML to infer the effects of device characteristics on system performance in inform optimization. At the integration level, we will explore mathematically provable scheduling and binding strategies targeting throughput and energy to co-optimize 3D integration of neuromorphic devices from thrusts 1 and 2. At the algorithm level, we will perform co-design with hardware and software to enable lifelong learning in neuromorphic computing.
