A programmable diffractive deep neural network based on a digital-coding metasurface array

Abstract

Artificial intelligence is facilitating human life in many aspects. Previous artificial intelligence has been mainly focused on computer algorithms (e.g. deep-learning and extremelearning) and integrated circuits. Recently, all-optical diffractive deep neural networks (D 2 NN) were realized by using passive structures, which can perform complicated functions designed by computer-based neural networks at the light speed. However, once a passive D 2 NN architecture is fabricated, its function will be fixed. Here, we propose a programmable artificial intelligence machine (PAIM) that can execute various intellectual tasks by realizing hierarchical connections of brain neurons via a multi-layer digital-coding metasurface array. Integrated with two amplifier chips in each meta-atom, its transmission coefficient covers a dynamic range of 35 dB (from -40 dB to -5 dB), which is the basis to construct the reprogrammable physical layers of D 2 NN, in which the digital meta-atoms make the artificial neurons alive. We experimentally show that PAIM can handle various deep-learning tasks for wave sensing, including image classifications, mobile communication coder-decoder, and real-time multi-beam focusing. In particular, we propose a reinforcement learning algorithm for on-site learning and discrete optimization algorithm for digital coding, making PAIM have autonomous intelligence ability and perform self-learning tasks without the support of extra computer.

An All-In-One Multifunctional Touch Sensor with Carbon-Based Gradient Resistance Elements

“…5 a). Based on the S-shaped AIOM touch sensor, a keystroke dynamics of biometric approach assisted by artificial neural network (ANN) was proposed [ 40 , 41 ]. ANN consisted of a neural network of interconnected processing neurons that mimicked the ability of the human brain to process complex linear and non-linear relationships using high-speed computing capabilities.…”

Near-field chiral excitation of universal spin-momentum locking transport of edge waves in microwave metamaterials

“…The future direction of spin-sorting metamaterials could be in the hybrid form of massive loading devices and novel structures. Related embedding fabrication and controlling technologies have shown significant achievements in recent digital coding metasurfaces 64,65 and will be transferred to the field of spin-sorting metamaterials. Combining the device-assisted metamaterials and controllable feeding circuits, we can tailor metamaterial waveguides and the source carrying angular momentum simultaneously.…”

Performing calculus with epsilon-near-zero metamaterials

“…In recent years, a surge in demand for high-speed computing has motivated numerous researchers to investigate the possibility of performing mathematical operations with photons instead of electrons, i.e., the concept of analog optical computing, to find an alternative path to overcome the speed and efficiency bottlenecks suffered by digital computers (11)(12)(13). Among these studies, some researchers aim at designing all-optical neuromorphic computing platforms by using the deep diffractive neural network (D 2 NN) frame (14)(15)(16)(17)(18), the Mach-Zehnder interferometer (MZI) array for chip-scale artificial neural networks (19,20), and memristors (21). Several photonic processors have achieved competitive performance with state-of-the-art electronic devices by harnessing phase-change-material memory arrays (22) or fiber dispersion (23), demonstrating the superiority of optical computing in speed and efficiency.…”