Application-specific optical processors have been considered disruptive technologies for modern computing that can fundamentally accelerate the development of artificial intelligence (AI) by offering substantially improved computing performance. Recent advancements in optical neural network architectures for neural information processing have been applied to perform various machine learning tasks. However, the existing architectures have limited complexity and performance; and each of them requires its own dedicated design that cannot be reconfigured to switch between different neural network models for different applications after deployment. Here, we propose an optoelectronic reconfigurable computing paradigm by constructing a diffractive processing unit (DPU) that can efficiently support different neural networks and achieve a high model complexity with millions of neurons. It allocates almost all of its computational operations optically and achieves extremely high speed of data modulation and large-scale network parameter updating by dynamically programming optical modulators and photodetectors. We demonstrated the reconfiguration of the DPU to implement various diffractive feedforward and recurrent neural networks and developed a novel adaptive training approach to circumvent the system imperfections. We applied the trained networks for high-speed classifying of handwritten digit images and human action videos over benchmark datasets, and the experimental results revealed a comparable classification accuracy to the electronic computing approaches. Furthermore, our prototype system built with off-the-shelf optoelectronic components surpasses the performance of state-of-the-art graphics processing units (GPUs) by several times on computing speed and more than an order of magnitude on system energy efficiency. We believe the proposed reconfigurable DPU is a remarkable step towards high-performance neuromorphic optoelectronic computing processors that can achieve real-time dynamic architecture configurations according to software and will facilitate a broad range of AI applications, e.g., autonomous driving, robotics, and edge computing.Computing processors driven by electronics have evolved dramatically over the past decade, from general-purpose central processing units (CPUs) 1 to custom computing platforms, e.g., GPUs 2 , FPGAs 3 , and ASICs 4,5 , to meet the ubiquitously increasing demand of computing resources. The progress of these silicon computing hardware platforms has greatly contributed to the resurgence of artificial intelligence (AI) by allowing the training of larger-scale and more complicated models 6,7 . We have witnessed the extensive applications of various neural computing architectures, e.g., convolutional neural networks (CNNs) 2,7 , recurrent neural networks (RNNs) 8 , spiking neural networks (SNNs) 9 , and reservoir computing (RC) 10 , in a broad range of fields. However, electronic hardware implementations have reached unsustainable performance growth as the exponential scaling of electr...

Large-scale single-cell analyses have become increasingly important given the role of cellular heterogeneity in complex biological systems. However, no current techniques enable optical imaging of uniquely-tagged individual cells. Fluorescence-based approaches can only distinguish a small number of distinct cells or cell groups at a time because of spectral crosstalk between conventional fluorophores. Here we investigate large-scale cell tracking using intracellular laser particles as imaging probes that emit coherent laser light with a characteristic wavelength. Made of silica-coated semiconductor microcavities, these laser particles have single-mode emission over a broad range from 1170 to 1580 nm with sub-nm linewidths, enabling massive spectral Users may view, print, copy, and download text and data-mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use:

Training an artificial neural network with backpropagation algorithms to perform advanced machine learning tasks requires an extensive computational process. This paper proposes to implement the backpropagation algorithm optically for in situ training of both linear and nonlinear diffractive optical neural networks, which enables the acceleration of training speed and improvement in energy efficiency on core computing modules. We demonstrate that the gradient of a loss function with respect to the weights of diffractive layers can be accurately calculated by measuring the forward and backward propagated optical fields based on light reciprocity and phase conjunction principles. The diffractive modulation weights are updated by programming a high-speed spatial light modulator to minimize the error between prediction and target output and perform inference tasks at the speed of light. We numerically validate the effectiveness of our approach on simulated networks for various applications. The proposed in situ optical learning architecture achieves accuracy comparable to in silico training with an electronic computer on the tasks of object classification and matrix-vector multiplication, which further allows the diffractive optical neural network to adapt to system imperfections. Also, the self-adaptive property of our approach facilitates the novel application of the network for all-optical imaging through scattering media. The proposed approach paves the way for robust implementation of large-scale diffractive neural networks to perform distinctive tasks all-optically.

No abstract

In this letter, we report the first study of the molecular relaxation dynamics in the glass-rubber transition region in polyisobutylene by 2D correlation dielectric relaxation spectroscopy (2DC-DRS). With the help of the high resolution and high sensitivity of 2DC-DRS, it is also the first time to observe and locate the positions of the Rouse modes and sub-Rouse modes in type-B polymers in the dielectric spectrum. 2DC-DRS was also applied to compare the temperature dependences of different molecular motions. Moreover, 2DC-DRS has been demonstrated as a powerful tool for studying the molecular motions with different time/length scales.

No abstract

The pixel size of a charge-coupled device (CCD) camera plays a major role in the image resolution, and the square pixels are attributed to the physical anisotropy of the sampling frequency. We synthesize the high sampling frequency directions from multiple frames acquired with different angles to enhance the resolution by 1.4× over conventional CCD orthogonal sampling. To directly demonstrate the improvement of frequency-domain diagonal extension (FDDE) microscopy, lens-free microscopy is used, as its resolution is dominantly determined by the pixel size. We demonstrate the resolution enhancement with a mouse skin histological specimen and a clinical blood smear sample. Further, FDDE is extended to lens-based photography with an ISO 12233 resolution target. This method paves a new way for enhancing the image resolution for a variety of imaging techniques in which the resolution is primarily limited by the sampling pixel size, for example, microscopy, photography, and spectroscopy.

High-speed visualization of three-dimensional (3D) processes across a large field of view with cellular resolution is essential for understanding living systems. Light-field microscopy (LFM) has emerged as a powerful tool for fast volumetric imaging. However, one inherent limitation of LFM is that the achievable lateral resolution degrades rapidly with the increase of the distance from the focal plane, which hinders the applications in observing thick samples. Here, we propose Spherical-Aberration-assisted scanning LFM (SAsLFM), a hardware-modification-free method that modulates the phase-space point-spread-functions (PSFs) to extend the effective high-resolution range along the z-axis by ~ 3 times. By transferring the foci to different depths, we take full advantage of the redundant light-field data to preserve finer details over an extended depth range and reduce artifacts near the original focal plane. Experiments on a USAF-resolution chart and zebrafish vasculatures were conducted to verify the effectiveness of the method. We further investigated the capability of SAsLFM in dynamic samples by imaging large-scale calcium transients in the mouse brain, tracking freely-moving jellyfish, and recording the development of Drosophila embryos. In addition, combined with deep-learning approaches, we accelerated the three-dimensional reconstruction of SAsLFM by three orders of magnitude. Our method is compatible with various phase-space imaging techniques without increasing system complexity and can facilitate high-speed large-scale volumetric imaging in thick samples.