Abstract:We introduce an all-optical Diffractive Deep Neural Network (D2NN) architecture that can learn to implement various functions after deep learning-based design of passive diffractive layers that work collectively. We experimentally demonstrated the success of this framework by creating 3D-printed D2NNs that learned to implement handwritten digit classification and the function of an imaging lens at terahertz spectrum. With the existing plethora of 3D-printing and other lithographic fabrication methods as well as spatial-light-modulators, this all-optical deep learning framework can perform, at the speed of light, various complex functions that computer-based neural networks can implement, and will find applications in all-optical image analysis, feature detection and object classification, also enabling new camera designs and optical components that can learn to perform unique tasks using D2NNs.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to develop a brand - new optical machine - learning framework, namely the all - optical diffractive deep neural network (Diffractive Deep Neural Network, D2NN). This framework utilizes the physical structures of multi - layer diffractive surfaces to achieve various functions, which can be learned and executed by passive diffractive layers designed by deep - learning algorithms. Specifically, the paper experimentally demonstrates that the D2NN fabricated by 3D printing technology can achieve the functions of handwritten digit classification and imaging lenses, and these tasks are completed within the terahertz frequency spectrum. The key points of the paper are as follows: 1. **All - optical implementation**: The D2NN completely relies on light diffraction and passive optical components and can perform complex machine - learning tasks at the speed of light without the need for electronic computing devices. 2. **Versatility**: Through different designs, the D2NN can achieve multiple functions, such as image analysis, feature detection, and object classification. 3. **Efficiency and scalability**: The D2NN can be mass - produced by using 3D printing and other high - throughput micro - and nano - manufacturing methods. It has the characteristics of high efficiency and low power consumption and is suitable for large - scale applications. 4. **Innovative design**: The design of the D2NN allows phase modulation at each node of the network and is optically interfered to connect to the nodes of the next layer, forming a unique network architecture. In conclusion, this paper aims to explore a new all - optical machine - learning method to physically implement a deep neural network in order to achieve high - speed, high - efficiency, and multi - functional optical computing.

All-Optical Machine Learning Using Diffractive Deep Neural Networks

Optical Analog Computing Devices Designed by Deep Neural Network

All-Optical Signal Processing of Vortex Beams with Diffractive Deep Neural Networks

Deep Learning-designed Diffractive Neural Networks

Advances and Progress of Diffractive Deep Neural Networks

Diffractive Deep Neural Networks at Visible Wavelengths

Ensemble learning of diffractive optical networks

Scale-, shift- and rotation-invariant diffractive optical networks

Fourier-space Diffractive Deep Neural Network.

Misalignment resilient diffractive optical networks

Residual D2NN: Training Diffractive Deep Neural Networks Via Learnable Light Shortcuts.

Design of task-specific optical systems using broadband diffractive neural networks

Time-lapse image classification using a diffractive neural network

Efficient Design Optimization for Diffractive Deep Neural Networks

All-optical autoencoder machine learning framework using diffractive processors

In Situ Optical Backpropagation Training of Diffractive Optical Neural Networks

<i>In situ</i> optical backpropagation training of diffractive optical neural networks

Scientific Computing with Diffractive Optical Neural Networks

1-bit Quantized On-chip Hybrid Diffraction Neural Network Enabled by Authentic All-optical Fully-connected Architecture

Class-specific Differential Detection in Diffractive Optical Neural Networks Improves Inference Accuracy

Optical authentication scheme based on all-optical neural network