All-optical information-processing capacity of diffractive surfaces

Abstract

The precise engineering of materials and surfaces has been at the heart of some of the recent advances in optics and photonics. These advances related to the engineering of materials with new functionalities have also opened up exciting avenues for designing trainable surfaces that can perform computation and machine-learning tasks through light–matter interactions and diffraction. Here, we analyze the information-processing capacity of coherent optical networks formed by diffractive surfaces that are trained to perform an all-optical computational task between a given input and output field-of-view. We show that the dimensionality of the all-optical solution space covering the complex-valued transformations between the input and output fields-of-view is linearly proportional to the number of diffractive surfaces within the optical network, up to a limit that is dictated by the extent of the input and output fields-of-view. Deeper diffractive networks that are composed of larger numbers of trainable surfaces can cover a higher-dimensional subspace of the complex-valued linear transformations between a larger input field-of-view and a larger output field-of-view and exhibit depth advantages in terms of their statistical inference, learning, and generalization capabilities for different image classification tasks when compared with a single trainable diffractive surface. These analyses and conclusions are broadly applicable to various forms of diffractive surfaces, including, e.g., plasmonic and/or dielectric-based metasurfaces and flat optics, which can be used to form all-optical processors.

Photonic matrix multiplication lights up photonic accelerator and beyond

“…Then, a modified D 2 NN based on class-specific differential detection was designed to improve the inference accuracy 47 . The information processing capacity of MPLC was recently discussed in detail by Kulce et al 121 , proving that the dimensionality of the all-optical solution space is linearly proportional to the number of phase planes. While it may be difficult to train the D 2 NN due to the existence of vanishing gradients, it has been suggested to address this issue by directly connecting the input and output using a learnable light shortcut, which offers a direct path for gradient backpropagation in training 122 .…”

Computational imaging without a computer: seeing through random diffusers at the speed of light

“…Some of the more recent work on imaging through diffusers has also focused on using deep learning methods to digitally recover the images of unknown objects [11,12,48,49]. Deep learning has been re-defining the state-of-the-art across many areas in optics, including optical microscopy [50][51][52][53][54][55], holography [56][57][58][59][60][61], inverse design of optical devices [62][63][64][65][66][67], optical computation and statistical inference [68][69][70][71][72][73][74][75][76][77], among others [78][79][80].…”

Metasurface-enabled on-chip multiplexed diffractive neural networks in the visible

“…As a kind of ONNs, the all-optical diffractive neural networks have been proposed and experimentally demonstrated by constructing 3D printing diffractive surfaces to form a physical network 21 at terahertz wavelengths and achieve specific functions 22 – 26 . Although no nonlinear activation function is applied, such multi-layer diffractive networks still exhibit a “depth” feature, i.e., the dimensionality of the transformation solution space is linearly proportional to the number of diffractive surfaces 27 . Nevertheless, the existing diffractive neural network devices, like conventional neural networks, cannot perform multiplexed information processing 28 – 31 .…”