A5089821617- Neural Networks and Reservoir Computing
- Photonic and Optical Devices
- Optical Network Technologies
- Advanced Fiber Laser Technologies
- Quantum Information and Cryptography
- Quantum Computing Algorithms and Architecture
- Laser-Matter Interactions and Applications
- Mechanical and Optical Resonators
- Neural Networks and Applications
- Quantum optics and atomic interactions
- Advanced Memory and Neural Computing
- Advanced Optical Sensing Technologies
- Semiconductor Lasers and Optical Devices
- Semiconductor Quantum Structures and Devices
- Solid State Laser Technologies
- Photorefractive and Nonlinear Optics
- Photonic Crystals and Applications
- Quantum and electron transport phenomena
- Adaptive optics and wavefront sensing
- Astronomy and Astrophysical Research
- Solar and Space Plasma Dynamics
- Optical Coatings and Gratings
- Spectroscopy Techniques in Biomedical and Chemical Research
- Nonlinear Optical Materials Studies
- Black Holes and Theoretical Physics
Massachusetts Institute of Technology
2019-2024
Vassar College
2018-2024
NTT Basic Research Laboratories
2024
Cambridge Electronics (United States)
2017-2024
NTT (United States)
2020-2023
Stanford University
2012-2022
Corning (United States)
2022
University of Central Florida
2022
IIT@MIT
2019-2021
National Institute of Informatics
2017-2019
Unconventional, special-purpose machines may aid in accelerating the solution of some hardest problems computing, such as large-scale combinatorial optimizations, by exploiting different operating mechanisms than those standard digital computers. We present a scalable optical processor with electronic feedback that can be realized at large scale room-temperature technology. Our prototype machine is able to find exact solutions of, or sample good approximate to, variety hard instances Ising...
Benchmarking the coherent Ising machine and D-Wave quantum annealer sheds light on importance of connectivity.
Recent success in deep neural networks has generated strong interest hardware accelerators to improve speed and energy consumption. This paper presents a new type of photonic accelerator based on coherent detection that is scalable large ($N \gtrsim 10^6$) can be operated at high (GHz) speeds very low (sub-aJ) energies per multiply-and-accumulate (MAC), using the massive spatial multiplexing enabled by standard free-space optical components. In contrast previous approaches, both weights...
Programmable photonic circuits of reconfigurable interferometers can be used to implement arbitrary operations on optical modes, facilitating a flexible platform for accelerating tasks in quantum simulation, signal processing, and artificial intelligence. A major obstacle scaling up these systems is static fabrication error, where small component errors within each device accrue produce significant the circuit computation. Mitigating this error usually requires numerical optimization...
Advanced machine learning models are currently impossible to run on edge devices such as smart sensors and unmanned aerial vehicles owing constraints power, processing, memory. We introduce an approach inference based delocalized analog processing across networks. In this approach, named Netcast, cloud-based "smart transceivers" stream weight data devices, enabling ultraefficient photonic inference. demonstrate image recognition at ultralow optical energy of 40 attojoules per multiply (<1...
We model the cooling of open optical and optomechanical resonators via feedback in linear quadratic Gaussian setting stochastic control theory. show that coherent schemes, which resonator is embedded an interferometer to achieve all-optical feedback, can outperform best possible measurement-based schemes quantum regime low steady-state excitation number. Such performance gains are attributed controller's ability process noncommuting output field quadratures simultaneously without loss...
Conventional computing architectures have no known efficient algorithms for combinatorial optimization tasks such as the Ising problem, which requires finding ground state spin configuration of an arbitrary graph. Physical machines recently been developed alternative to conventional exact and heuristic solvers; however, these typically suffer from decreased convergence probability or universality high edge-density graphs graph weights, respectively. We experimentally demonstrate a...
Analog optical and electronic hardware has emerged as a promising alternative to digital electronics improve the efficiency of deep neural networks (DNNs). However, previous work been limited in scalability (input vector length
Realization of a room-temperature ultra-fast photon-number-resolving (PNR) quantum nondemolition (QND) measurement would have significant implications for photonic information processing (QIP), enabling, e.g., deterministic computation in discrete-variable architectures, but the requirement strong coupling has hampered development scalable implementations. In this work, we propose and analyze nonlinear-optical route to PNR QND using quadratic (i.e., $\chi^{(2)}$) nonlinear interactions. We...
We identify and demonstrate a regime of operation in optical parametric oscillators (OPOs) which the formation temporal simultons produces stable half-harmonic pulses. Simultons are simultanous bright-dark solitons signal field at frequency $\omega$ pump $2\omega$ form quadratic nonlinear medium. The an OPO is evidenced by sech$^2$ spectra with broad instantaneous bandwidths increase power, large slope efficiencies. In contrast to conventional synchronously pumped OPOs, this achieved using...
Abstract As deep neural network (DNN) models grow ever-larger, they can achieve higher accuracy and solve more complex problems. This trend has been enabled by an increase in available compute power; however, efforts to continue scale electronic processors are impeded the costs of communication, thermal management, power delivery clocking. To improve scalability, we propose a digital optical (DONN) with intralayer interconnects reconfigurable input values. The path-length-independence energy...
As deep neural networks (DNNs) revolutionize machine learning, energy consumption and throughput are emerging as fundamental limitations of CMOS electronics. This has motivated a search for new hardware architectures optimized artificial intelligence, such electronic systolic arrays, memristor crossbar optical accelerators. Optical systems can perform linear matrix operations at exceptionally high rate efficiency, motivating recent demonstrations low latency algebra below photon per...
Multiport interferometers based on integrated beamsplitter meshes are widely used in photonic technologies. While the rectangular mesh is favored for its compactness and uniformity, geometry resists conventional self-configuration approaches, which essential to programming large presence of fabrication error. Here, we present a configuration algorithm, related $2\ifmmode\times\else\texttimes\fi{}2$ block decomposition unitary matrix, that overcomes this limitation. Our proposed algorithm...
Component errors limit the scaling of programmable coherent photonic circuits. These arise because standard tunable coupler-the Mach-Zehnder interferometer (MZI)-cannot be perfectly programmed to cross state. Here, we introduce two modified circuit architectures that overcome this limitation: (1) a 3-splitter MZI mesh for generic errors, and (2) broadband MZI+Crossing design correlated errors. Because these designs allow perfect realization state, matrix fidelity no longer degrades with...
Photonic integrated circuits with second-order ( χ (2) ) nonlinearities are rapidly scaling to remarkably low powers. At this time, state-of-the-art devices achieve saturated nonlinear interactions thousands of photons when driven by continuous-wave lasers, and further reductions in these energy requirements enabled the use ultrafast pulses may soon push optics into realm single-photon nonlinearities. This tutorial reviews recent developments photonics, discusses design strategies for...
Over the last few decades, nonlinear optics has become significantly more nonlinear, traversing nearly a billionfold improvement in energy efficiency, with ultrafast nanophotonics particular emerging as frontier for combining both spatial and temporal engineering. At present, cutting-edge experiments place us just above mesoscopic regime, where hundred photons suffice to trigger highly dynamics. In contrast classical or deep-quantum optics, mesoscale is characterized by dynamical...
Abstract The development of physical simulators, called Ising machines, that sample from low energy states the Hamiltonian has potential to transform our ability understand and control complex systems. However, most implementations such machines have been based on a similar concept is closely related relaxational dynamics as in simulated, mean-field, chaotic, quantum annealing. Here we show includes nonrelaxational component associated with finite positive Gibbs entropy production rate can...
Realistic multiport interferometers (beamsplitter meshes) are sensitive to component imperfections, and this sensitivity increases with size. Self-configuration techniques can be employed correct these but not all equal. This paper highlights the importance of algorithmic stability in self-configuration. Na\"{\i}ve approaches based on sequentially setting matrix elements unstable perform poorly for large meshes, while power ratios well cases, even presence errors. Based insight, we propose a...
While analog neural network (NN) accelerators promise massive energy and time savings, an important challenge is to make them robust static fabrication error. Present-day training methods for programmable photonic interferometer circuits, a leading NN platform, do not produce networks that perform well in the presence of hardware errors. Moreover, existing error correction techniques either require individual retraining every (which impractical edge setting with millions devices), place...
We study the cooling performance of optical-feedback controllers for open optical and mechanical resonators in Linear Quadratic Gaussian setting stochastic control theory. utilize analysis numerical optimization closed-loop models based on quantum differential equations to show that coherent schemes, where we embed resonator an interferometer achieve all-optical feedback, can outperform optimal measurement-based feedback schemes regime low steady-state excitation number. These gains are...
Many tasks in our modern life, such as planning an efficient travel, image processing and optimizing integrated circuit design, are modeled complex combinatorial optimization problems with binary variables. Such can be mapped to finding a ground state of the Ising Hamiltonian, thus various physical systems have been studied emulate solve this problem. Recently, networks mutually injected optical oscillators, called coherent machines, developed promising solvers for problem, benefiting from...
Ongoing advances in semiconductor fabrication are expected to enable nonlinear optical circuits operate at extremely low switching energies, where quantum effects become important. This work describes semiclassical simulations study the of noise large digital logic containing hundreds components. The authors find that amplitudes fluctuations do not increase as signals propagate through circuit, which is promising for scaling up devices.