Quantum Sequence Recognition Via Hybrid Analog Spiking-Neuron-Like Elements
Abstract
We introduce a wholly novel circuit-design architecture for creating ~Quantum Analog Hybrid (QAH)~ computers that can learn and recognize arbitrary time-varying sequences. Our approach exploits the power of Quantum computing with Analog spiking neuron-like oscillatory circuit elements that interact via Hybrid analog-digital signals. We show that the QAH method scales to arbitrary complexity, unlike all current competing approaches (quantum, classical, neural, and other hybrids), for the first time enabling quantum methods to be applied to large-scale sequence signals.Current recognition systems are very ill suited for sequence processing, typically requiring RNN adaptations that are expensive to build, requiring extensive hand-tailoring, as well as being expensive to execute, with poor scaling characteristics for large-scale sequence data. Methods are needed for massively parallel computation that can provide scalable bases for fieldable devices for on-site recognition of sequential data.Sequential time-varying signals (such as RF) are ubiquitous; the proposed QAHoutperformance of current analysis methods will greatly improve the ability to capture and decode such signals with fieldable mobile devices, with a corresponding impact on activities that are dependent on these real-time streams.All extant attempts to develop parallel hardware (whether quantum or non-quantum neurallike systems) suffer from scale-up limitations; although proofs of principle have been produced, no large quantum systems capable of substantial data throughput yet exist. The proposed QAH system ~ a hybrid scheme with moderate-precision analog spiking neurons that collectively interact ~ scales to arbitrary complexity or arbitrary precision just as digital computers do, while exhibiting significantly more power than existing computers.The proposed work relies on three crucial findings (Emulation of Quantum Dynamics via Classical Circuits, Analog Circuits, and a Hybrid collective architecture), which, taken together, establish the platform for the QAH method. Each of these three critical constituent elements already rests on a demonstrated set of findings; these will be definitively tested early in the proposed timeline.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 13, 2019
- Source ID
- N000141912434
Entities
People
- Rahul Sarpeshkar
Organizations
- Board of Trustees of Dartmouth College
- Office of Naval Research
- United States Navy