NEB: Physics-Inspired Non-Boolean Computation based on Spatial- Temporal Wave Excitations

点击翻译按钮获取中文摘要

This project is awarded under the Nanoelectronics for 2020 and Beyond competition, with support by multiple Directorates and Divisions at the National Science Foundation as well as by the Nanoelectronics Research Initiative of the Semiconductor Research Corporation. This work is motivated by the vision of future computer chips with millions of cores, where each core has a complexity similar to early microprocessors. Furthermore, it is envisioned that computation in such systems might be done in an entirely new way, more akin to wave propagation in a brain-like network of processing elements. This work will identify wave-like non-Boolean computational primitives based on the spatial-temporal characteristics of such processor networks, and inspiration will be gained from known wave behavior in physics and applications derived from these. In addition, nanoelectronic systems will be explored where such wave-type excitations might find a natural implementation. This research will identify, on the physical level, computational building blocks based on nanoelectronic structures with spatial-temporal wave-like dynamics, and on the computer science level, ways to perform computation with such non-Boolean primitives. This interdisciplinary research, which spans the whole range from nanostructures to new paradigms for computing, will require a team with expertise that spans the whole range from physics to computer science. This proposal will lay the groundwork for a radically different approach to information processing, which is based on physics-inspired and brain-like wave behavior in large-scale arrays of nanoelectronic processing elements. Specifically, our research will identify computational building blocks of future computing systems, along with the inherent state variables, where each computational task directly maps onto the underlying physical structure. This project will also identify which information-processing tasks will find their natural implementation in such architectures. This proposed research is planned in synergy with the Semiconductor Research Corporation (SRI) Nanoelectronics Research Initiative (NRI) and activities at the SRC-NRI Midwest Institute for Nanoelectronics Discovery (MIND). In particular, this research will partner with Dr. George Bourianoff at Intel, who leads a study group on Ultra Low Power Application Specific Non-Boolean Systems, which is closely related to the research proposed here.If successful, this work would impact the way information processing is done in future information-processing systems. This work will not merely lead to incremental improvements in information processing systems, but open the door to an entirely new approach to computer architecture. Computation no longer will be based on elementary Boolean logic functions, but on physics-inspired and brain-like wave activity. This will necessitate a complete re-thinking of how computation is being done. As part of this research, graduate students will be educated and trained in this new way of thinking about how computation is done in a more brain-like fashion. Building upon active REU and RET programs, undergraduates and teachers will participate in this work. It is a stated goal to perform this research in an environment of maximum inclusivity of underrepresented groups.