Smaller matrix multiplication for less complex inference tasks are still challenged by a non-negligible latency predominantly due to the access overhead of the various memory hierarchies and the latency in executing each instruction in the GPU. Given this context of computational hardware for obtaining architectures, it is necessary to explore and reinvent the operational paradigms of current computing platforms when performing matrix algebra, by replacing sequential and temporized operations to memory, with massively parallelized distributed analog dynamical units, towards delivering efficient post-CMOS devices and systems summarized as non-von Neumann architectures. Within this paradigm shift the wave nature of light and related inherent operations, such as interference and diffraction, enable seamless mathematical operations (e.g. multiplication) to be executed in the optical domain. Hence, photonic processors can play a major role in enhancing computational throughput and concurrently reducing the power consumption of neuromorphic platform. Future technologies should perform computing tasks in the domain in which their time varying input signals lay, thus exploiting and leveraging their intrinsic physical operations. There are 3 key innovations that we are considering combining in this PTC AI system, these include the (1) PTC architecture itself, which allows for a modular and scale design and incorporates multiplexing schemes such as WDM offered in photonics [Sorger Appl. Phys. Rev. 2020], (2) Photonic nonvolatile but programmable memory (P-RAM), allowing to re-write the B-matrix of the VMM or, exemplary, the kernel of a CNN. The nonvolatility and retention of state in the photonic system allows for compute-in-memory functionality, thus bypassing the memory-access bottlenecks known in Van-Neumann architectures. And (3) If the data entering the PTC happens to be electronic and digital, then DAC is needed such as a photonic parallel binary-weighted DAC. Potential NASA Applications (Limit 1500 characters, approximately 150 words): The Cognitive Communications Project, through the Human Exploration and Operations Mission Directorate (HEOMD) Space Communications and Navigation (SCaN) Program, is one potential customer of this to-be-developed PTC ML system. Potential Non-NASA Applications (Limit 1500 characters, approximately 150 words): Dual use is for DOD and National Security applications (NSA) such as deep-surveillance, UAV navigation/ranging and data pre-processing at the edge of the network. The global IT market is $5.9T (data 2018), whereas photonics technologies capture about 10%. Since Optelligence LLC is in R&D, the average R&D investment fraction is ~15% ($~100B), with an estimated U.S. share of about 25% ($25B). Duration: 6
