Neural networks have enabled applications in artificial intelligence through machine learning, and neuromorphic computing. Software implementations of neural networks on conventional computers that have separate memory and processor (and that operate
sequentially) are limited in speed and energy efficiency. Neuromorphic engineering aims to build processors in which hardware mimics neurons and synapses in the brain for distributed and parallel processing. Neuromorphic engineering enabled by photonics (optical physics) can offer sub-nanosecond latencies and high bandwidth with low energies to extend the domain of artificial intelligence and neuromorphic computing applications to machine learning acceleration, nonlinear programming, intelligent signal processing, etc. Photonic neural networks have been demonstrated on integrated platforms and free-space optics depending on the class of applications being targeted. Here, we discuss the prospects and demonstrated applications of these photonic neural networks.
While information is ubiquitously generated, shared, and analyzed in a modern-day life, there is still some controversy around the ways to asses the amount and quality of information inside a noisy optical channel. A number of theoretical approaches
based on, e.g., conditional Shannon entropy and Fisher information have been developed, along with some experimental validations. Some of these approaches are limited to a certain alphabet, while others tend to fall short when considering optical beams with non-trivial structure, such as Hermite-Gauss, Laguerre-Gauss and other modes with non-trivial structure. Here, we propose a new definition of classical Shannon information via the Wigner distribution function, while respecting the Heisenberg inequality. Following this definition, we calculate the amount of information in a Gaussian, Hermite-Gaussian, and Laguerre-Gaussian laser modes in juxtaposition and experimentally validate it by reconstruction of the Wigner distribution function from the intensity distribution of structured laser beams. We experimentally demonstrate the technique that allows to infer field structure of the laser beams in singular optics to assess the amount of contained information. Given the generality, this approach of defining information via analyzing the beam complexity is applicable to laser modes of any topology that can be described by well-behaved functions. Classical Shannon information defined in this way is detached from a particular alphabet, i.e. communication scheme, and scales with the structural complexity of the system. Such a synergy between the Wigner distribution function encompassing the information in both real and reciprocal space, and information being a measure of disorder, can contribute into future coherent detection algorithms and remote sensing.
Low latency, high throughput inference on Convolution Neural Networks (CNNs) remains a challenge, especially for applications requiring large input or large kernel sizes. 4F optics provides a solution to accelerate CNNs by converting convolutions int
o Fourier-domain point-wise multiplications that are computationally free in optical domain. However, existing 4F CNN systems suffer from the all-positive sensor readout issue which makes the implementation of a multi-channel, multi-layer CNN not scalable or even impractical. In this paper we propose a simple channel tiling scheme for 4F CNN systems that utilizes the high resolution of 4F system to perform channel summation inherently in optical domain before sensor detection, so the outputs of different channels can be correctly accumulated. Compared to state of the art, channel tiling gives similar accuracy, significantly better robustness to sensing quantization (33% improvement in required sensing precision) error and noise (10dB reduction in tolerable sensing noise), 0.5X total filters required, 10-50X+ throughput improvement and as much as 3X reduction in required output camera resolution/bandwidth. Not requiring any additional optical hardware, the proposed channel tiling approach addresses an important throughput and precision bottleneck of high-speed, massively-parallel optical 4F computing systems.
Machine-intelligence has become a driving factor in modern society. However, its demand outpaces the underlying electronic technology due to limitations given by fundamental physics such as capacitive charging of wires, but also by system architectur
e of storing and handling data, both driving recent trends towards processor heterogeneity. Here we introduce a novel amplitude-only Fourier-optical processor paradigm capable of processing large-scale ~(1,000 x 1,000) matrices in a single time-step and 100 microsecond-short latency. Conceptually, the information-flow direction is orthogonal to the two-dimensional programmable-network, which leverages 10^6-parallel channels of display technology, and enables a prototype demonstration performing convolutions as pixel-wise multiplications in the Fourier domain reaching peta operations per second throughputs. The required real-to-Fourier domain transformations are performed passively by optical lenses at zero-static power. We exemplary realize a convolutional neural network (CNN) performing classification tasks on 2-Megapixel large matrices at 10 kHz rates, which latency-outperforms current GPU and phase-based display technology by one and two orders of magnitude, respectively. Training this optical convolutional layer on image classification tasks and utilizing it in a hybrid optical-electronic CNN, shows classification accuracy of 98% (MNIST) and 54% (CIFAR-10). Interestingly, the amplitude-only CNN is inherently robust against coherence noise in contrast to phase-based paradigms and features an over 2 orders of magnitude lower delay than liquid crystal-based systems. Beyond contributing to novel accelerator technology, scientifically this amplitude-only massively-parallel optical compute-paradigm can be far-reaching as it de-validates the assumption that phase-information outweighs amplitude in optical processors for machine-intelligence.
Analog photonic solutions offer unique opportunities to address complex computational tasks with unprecedented performance in terms of energy dissipation and speeds, overcoming current limitations of modern computing architectures based on electron f
lows and digital approaches. The lack of modularization and lumped element reconfigurability in photonics has prevented the transition to an all-optical analog computing platform. Here, we explore a nanophotonic platform based on epsilon-near-zero materials capable of solving in the analog domain partial differential equations (PDE). Wavelength stretching in zero-index media enables highly nonlocal interactions within the board based on the conduction of electric displacement, which can be monitored to extract the solution of a broad class of PDE problems. By exploiting control of deposition technique through process parameters, we demonstrate the possibility of implementing the proposed nano-optic processor using CMOS-compatible indium-tin-oxide, whose optical properties can be tuned by carrier injection to obtain programmability at high speeds and low energy requirements. Our nano-optical analog processor can be integrated at chip-scale, processing arbitrary inputs at the speed of light.
The interaction between excitons and phonons in semiconductor nanocrystals plays a crucial role in the exciton energy spectrum and dynamics, and thus in their optical properties. We investigate the exciton2 phonon coupling in giant-shell CdSe/CdS cor
e-shell nanocrystals via resonant Raman spectroscopy. The Huang-Rhys parameter is evaluated by the intensity ratio of the longitudinal-optical (LO) phonon of CdS with its first multiscattering (2LO) replica. We used four different excitation wavelengths in the range from the onset of the CdS shell absorption to well above the CdS shell band edge to get insight into resonance effects of the CdS LO phonon with high energy excitonic transitions. The isotropic spherical giant-shell nanocrystals show consistently stronger exciton-phonon coupling as compared to the anisotropic rod-shaped dot-in-rod (DiR) architecture, and the 2LO/LO intensity ratio decreases for excitation wavelengths approaching the CdS band edge. The strong exciton-phonon coupling in the spherical giant-shell nanocrystals can be related to the delocalization of the electronic wave functions. Furthermore, we observe the radial breathing modes of the GS nanocrystals and their overtones by ultralow frequency Raman spectroscopy with nonresonant excitation, using laser energies well below the band gap of the heteronanocrystals, and highlight the differences between higher order
Plasmonic metasurfaces have spawned the field of flat optics using nanostructured planar metallic or dielectric surfaces that can replace bulky optical elements and enhance the capabilities of traditional far-field optics. Furthermore, the potential
of flat optics can go far beyond far-field modulation, and can be exploited for functionality in the near-field itself. Here, we design metasurfaces based on aperiodic arrays of plasmonic Au nanostructures for tailoring the optical near-field in the visible and near-infrared spectral range. The basic element of the arrays is a rhomboid that is modulated in size, orientation and position to achieve the desired functionality of the micron-size metasurface structure. Using two-photon-photoluminescence as a tool to probethe near-field profiles in the plane of the metasurfaces, we demonstrate the molding of light into different near-field intensity patterns and active pattern control via the far-field illumination. Finite element method simulations reveal that the near-field modulation occurs via a combination of the plasmonic resonances of the rhomboids and field enhancement in the nanoscale gaps in between the elements. This approach enables optical elements that can switch the near-field distribution across the metasurface via wavelength and polarization of the incident far-field light, and provides pathways for light matter interaction in integrated devices.
Digital-to-analog converters (DAC) are indispensable functional units in signal processing instrumentation and wide-band telecommunication links for both civil and military applications. Since photonic systems are capable of high data throughput and
low latency, an increasingly found system limitation stems from the required domain-crossing such as digital-to-analog, and electronic-to-optical. A photonic DAC implementation, in contrast, enables a seamless signal conversion with respect to both energy efficiency and short signal delay, often require bulky discrete optical components and electric-optic transformation hence introducing inefficiencies. Here, we introduce a novel coherent parallel photonic DAC concept along with an experimental demonstration capable of performing this digital-to-analog conversion without optic-electric-optic domain crossing. This design hence guarantees a linear intensity weighting among bits operating at high sampling rates, yet at a reduced footprint and power consumption compared to other photonic alternatives. Importantly, this photonic DAC could create seamless interfaces of next-generation data processing hardware for data-centers, task-specific compute accelerators such as neuromorphic engines, and network edge processing applications.
When solving, modelling or reasoning about complex problems, it is usually convenient to use the knowledge of a parallel physical system for representing it. This is the case of lumped-circuit abstraction, which can be used for representing mechanica
l and acoustic systems, thermal and heat-diffusion problems and in general partial differential equations. Integrated photonic platforms hold the prospect to perform signal processing and analog computing inherently, by mapping into hardware specific operations which relies on the wave-nature of their signals, without trusting on logic gates and digital states like electronics. Although, the distributed nature of photonic platforms leads to the absence of an equivalent approximation to Kirchhoffs law, the main principle used for representing physical systems using circuits. Here we argue that in absence of a straightforward parallelism and homomorphism can be induced. We introduce a photonic platform capable of mimicking Kirchhoffs law in photonics and used as node of a finite difference mesh for solving partial differential equation using monochromatic light in the telecommunication wavelength. We experimentally demonstrate generating in one-shot discrete solutions of a Laplace partial differential equation, with an accuracy above 95% relative to commercial solvers, for an arbitrary set of boundary conditions. Our photonic engine can provide a route to achieve chip-scale, fast (10s of ps), and integrable reprogrammable accelerators for the next generation hybrid high performance computing.
Driven by machine-learning tasks neural networks have demonstrated useful capabilities as nonlinear hypothesis classifiers. The underlying technologies performing the dot product multiplication, the summation, and the nonlinear thresholding on the in
put data in electronics, however, are limited by the same capacitive challenges known from electronic integrated circuits. The optical domain, in contrast, provides low delay interconnectivity suitable for such node distributed non Von Neumann architectures relying on dense node to node communication. Thus, once the neural networks weights are set, the delay of the network is just given by the time of flight of the photon, which is in the picosecond range for photonic integrated circuits. However, the functionality of memory for storing the trained weights does not exists in optics, thus demanding a fresh look to explore synergies between photonics and electronics in neural networks. Here we provide a roadmap to pave the way for emerging hybridized photonic electronic neural networks by taking a detailed look into a single nodes perceptron, discussing how it can be realized in hybrid photonic electronic heterogeneous technologies. We show that a set of materials exist that exploit synergies with respect to a number of constrains including electronic contacts, memory functionality, electrooptic modulation, optical nonlinearity, and device packaging. We find that the material ITO, in particular, could provide a viable path for both the perceptron weights and the nonlinear activation function, while simultaneously being a foundry process near material. We finally identify a number of challenges that, if solved, could accelerate the adoption of such heterogeneous integration strategies of emerging memory materials into integrated photonics platforms for real time responsive neural networks.
