اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدAll-Optical Image Identification with Programmable Matrix Transformation
74
0
0.0
(
0
)
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
An optical neural network is proposed and demonstrated with programmable matrix transformation and nonlinear activation function of photodetection (square-law detection). Based on discrete phase-coherent spatial modes, the dimensionality of programmable optical matrix operations is 30~37, which is implemented by spatial light modulators. With this architecture, all-optical classification tasks of handwritten digits, objects and depth images are performed on the same platform with high accuracy. Due to the parallel nature of matrix multiplication, the processing speed of our proposed architecture is potentially as high as7.4T~74T FLOPs per second (with 10~100GHz detector)
قيم البحث
اقرأ أيضاً
Deeplearning algorithms are revolutionising many aspects of modern life. Typically, they are implemented in CMOS-based hardware with severely limited memory access times and inefficient data-routing. All-optical neural networks without any electro-optic
Measurement of the optical transmission matrix (TM) of an opaque material is an advanced form of space-variant aberration correction. Beyond imaging, TM-based methods are emerging in a range of fields including optical communications, optical micro-m
anipulation, and optical computing. In many cases the TM is very sensitive to perturbations in the configuration of the scattering medium it represents. Therefore applications often require an up-to-the-minute characterisation of the fragile TM, typically entailing hundreds to thousands of probe measurements. In this work we explore how these measurement requirements can be relaxed using the framework of compressive sensing: incorporation of prior information enables accurate estimation from fewer measurements than the dimensionality of the TM we aim to reconstruct. Examples of such priors include knowledge of a memory effect linking input and output fields, an approximate model of the optical system, or a recent but degraded TM measurement. We demonstrate this concept by reconstructing a full-size TM of a multimode fibre supporting 754 modes at compression ratios down to ~5% with good fidelity. The level of compression achievable is dependent upon the strength of our priors. We show in this case that imaging is still possible using TMs reconstructed at compression ratios down to ~1% (8 probe measurements). This compressive TM sampling strategy is quite general and may be applied to any form of scattering system about which we have some prior knowledge, including diffusers, thin layers of tissue, fibre optics of any known refractive profile, and reflections from opaque walls. These approaches offer a route to measurement of high-dimensional TMs quickly or with access to limited numbers of measurements.
All-optical binary convolution with a photonic spiking vertical-cavity surface-emitting laser (VCSEL) neuron is proposed and demonstrated experimentally for the first time. Optical inputs, extracted from digital images and temporally encoded using re
ctangular pulses, are injected in the VCSEL neuron which delivers the convolution result in the number of fast (<100 ps long) spikes fired. Experimental and numerical results show that binary convolution is achieved successfully with a single spiking VCSEL neuron and that all-optical binary convolution can be used to calculate image gradient magnitudes to detect edge features and separate vertical and horizontal components in source images. We also show that this all-optical spiking binary convolution system is robust to noise and can operate with high-resolution images. Additionally, the proposed system offers important advantages such as ultrafast speed, high energy efficiency and simple hardware implementation, highlighting the potentials of spiking photonic VCSEL neurons for high-speed neuromorphic image processing systems and future photonic spiking convolutional neural networks.
we investigate the transmission of probe laser beam in a coupled-cavity system with polaritons by using standard input-output relation of optical fields, and proposed a theoretical schema for realizing a polariton-based photonic transistor. On accoun
t of effects of exciton-photon coupling and single-photon optomechanical coupling, a probe laser field can be either amplified or attenuated by another pump laser field when it passes through a coupled-cavity system with polaritons. The Stokes and anti-Stokes scattered effect of output prober laser can also be modulated. Our results open up exciting possibilities for designing photonic transistors.
With this paper we bring about a discussion on the computing potential of complex optical networks and provide experimental demonstration that an optical fiber network can be used as an analog processor to calculate matrix inversion. A 3x3 matrix is
inverted as a proof-of-concept demonstration using a fiber network containing three nodes and operating at telecomm wavelength. For an NxN matrix, the overall solving time (including setting time of the matrix elements and calculation time of inversion) scales as O(N^2), whereas matrix inversion by most advanced computer algorithms requires ~O(N^2.37) computational time. For well-conditioned matrices, the error of the inversion performed optically is found to be less than 3%, limited by the accuracy of measurement equipment.
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
