With the continuous miniaturization of conventional integrated circuits, obstacles such as excessive cost, increased resistance to electronic motion, and increased energy consumption are gradually slowing down the development of electrical computing and constraining the application of deep learning. Optical neuromorphic computing presents various opportunities and challenges compared with the realm of electronics. Algorithms running on optical hardware have the potential to meet the growing computational demands of deep learning and artificial intelligence. Here, we review the development of optical neural networks and compare various research proposals. We focus on fiber-based neural networks. Finally, we describe some new research directions and challenges.

The refractive index (RI) measurment of the liquid often changes with the sensor's temperature. Accurate RI measurements require simultaneous measurements of the liquid temperature to calibrate the RI measurement. This study proposed what we believe to be a novel dual D-shaped plastic optical fiber sensor capable of simultaneously measuring both RI and temperature. A fluorescent material made of rhodamine B (RhB) is embedded in one of the D-shaped structures of the dual D-shaped fiber, which can be excited by a green laser to produce orange fluorescence. The fluorescence and the input laser of the fiber are superimposed at the output end of the fiber to form a specklegram containing information of two wavelengths. It was found through experiments that the effects of temperature and RI change on the red and green channels of the specklegrams are different, and the neural network can learn this feature to complete the measurement of both RI and temperature parameters. For RI sensing, the maximum error between the average predicted value and the true value of the test set is 0.0005. For temperature sensing, the maximum error between the average predicted value and the true value of the test set is 0.26°C. In addition, because the intensity change of the fluorescence varies linearly with RI and temperature, the sensor also has good stability.

Optical neural networks (ONNs) are getting more and more attention due to their advantages such as high-speed and low power consumption. However, in a non-ideal environment, the noise and low-bits control may heavily lead to a decrease in the accuracy of ONNs. Since there is AD/DA conversion in a simulated neural network, it needs to be quantified in the model. In this paper, we propose a quantitative method to adapt ONN to a non-ideal environment with fixed-point transmission, based on the new chip structure we designed previously. An MNIST hand-written data set was used to test and simulate the model we established. The experimental results showed that the quantization-noise model we established has a good performance, for which the accuracy was up to about 96%. Compared with the electrical method, the proposed quantization method can effectively solve the non-ideal ONN problem.

Gaze tracking technology has a wide range of applications in the fields of VR/AR glasses. In this paper, a plastic optical fiber (POF) is used as a light transmission waveguide to transmit the light reflected from the eye to the outside of the device for processing, thereby eliminating the need for in-device camera installations. By processing two fan-shaped surfaces on the POF at a 45° angle relative to the vertical direction, the POF gains the ability to couple light from the side. The reflected light of the eye can be transmitted to the outside of the device through the POF. The specklegram corresponding to 76 different gaze directions is classified by SE-Resnet18, and the accuracy reached 96.9%. The gaze tracking system is low cost and simple in structure and has potential application in fields such as AR glasses, human–computer interaction (HCI), and medical diagnosis.

Multiple-input multiple-output (MIMO) detectors have been a key technology in communication systems. In this paper, a new MIMO detector is designed by combining the adaptive learning rate (ALR) with the convolutional neural network (CNN) and successfully implementing it in a mode division multiplexing (MDM) optical transmission system. The results show that the training and test accuracy of the signal in the system we proposed reaches 100%. What is more, we used the ALR-CNN to compare the performance with conventional detection algorithms. The results confirm that our DLNN exceeds the conventional MIMO detectors in performance and is able to achieve the ideal QPSK BER level. The minimum difference in the SNR is about 9.5 dB at a BER of the 10⁻³ order.

There is an increasing demand for data with the development of the world, and various fiber optic multiplexing techniques have become an important research direction to improve transmission capacity. However, the transmitted signals are subject to great interference due to mode coupling and mode dispersion, which require multiple-input multiple-output (MIMO) digital signal processing techniques to restore the quality of the transmitted signals. In this paper, a novel MIMO detector is designed using an adaptive learning recurrent neural network and successfully implemented in a mixed wavelength-division-mode-division-multiplexing (WDM-MDM) optical transmission system, and its performance is compared with that of the forced-zero detector and the minimum-mean-square-error detector. The results show that the introduction of an adaptive machine learning model in MIMO detection for WDM-MDM optical transmission systems can significantly improve the quality of the transmitted signals and achieve better performance than other MIMO detection algorithms while maintaining a faster computational speed and a lower number of parameters.