Unsupervised Neural Network for Modulation Format Discrimination and Identification

In this study, we develop an intelligent constellation diagram monitoring method for fiber optic communications based on a small sample learning network. The method is designed to identify the modulation format, estimate the optical signal-to-noise ratio (OSNR), and predict the fiber transmission distance by analyzing a small amount of constellation diagram data. The model was trained using the C-way-K-shot strategy with the Adam optimization algorithm, with a learning rate of 0.001. In terms of modulation format identification, the model was tested under different OSNR conditions, and the results showed that under low, medium, and high OSNR conditions, the 5-way-3-shot task models achieved average accuracies over 10 iterations of 95.73%, 98.35%, and 97.86%, respectively. In contrast, for the 5-way-5-shot task, the accuracies increased to 98.38%, 99.23%, and 98.74%, respectively. This indicates that, even under low OSNR conditions, the small-sample learning network can achieve a high modulation format recognition rate, demonstrating the model’s fast adaptability and high accuracy.Meanwhile, the OSNR estimation in low-order modulation formats achieves a 100% recognition rate, and for high-order modulation formats, the estimation also performs well. The small-sample learning network demonstrates good performance even with a limited data sample size. The recognition rate of transmission distance estimation for low-order modulation formats, such as BPSK and QPSK, performs well under low, medium, and high OSNR conditions. However, performance in recognizing higher-order modulation formats, such as 64QAM, is poor, highlighting both the adaptability and limitations of the model when dealing with different modulation formats. This study demonstrates the potential of small-sample learning networks for intelligent constellation diagram monitoring in fiber optic communications, especially in terms of their high accuracy and robustness in modulation format identification, OSNR estimation, and fiber link transmission distance prediction. The method effectively utilizes a small sample size to achieve fast and accurate monitoring, thus providing effective technical support for the intelligent management and maintenance of fiber optic communication systems.

In coherent optical communication systems, where multiple modulation formats are mixed and variable, the correct identification of signal modulation formats provides the foundation for subsequent performance improvement using digital algorithms. A modulation format identification (MFI) scheme based on signal constellation diagrams and support vector machine (SVM) is proposed. Firstly, the signal constellation diagrams are divided by the fractal dimension of the weighted linear least squares (WLS-FD) algorithm, and the fractal dimension (FD) in each region is calculated, which is regarded as one of the image features. Then, the feature values of the image in different directions are extracted by the gray-level co-occurrence matrix (GLCM), and their mean and variance are calculated, which is regarded as another feature. Finally, the two features are input into the modulation format classifier constructed by the SVM to achieve MFI in coherent optical communication systems. To verify the feasibility and superiority of the scheme, we compare it with the MFI scheme based on higher-order statistical (HOS) features, GLCM features, and FD features, respectively. Further, we built a 30 GBaud coherent optical communication system with fiber lengths of 80 km and 120 km, where the optical signal-to-noise ratio (OSNR) ranges from 0 dB to 30 dB. The proposed MFI scheme identifies seven modulation formats: QPSK, 8QAM, 16QAM, 32QAM, 64QAM, 128QAM, and 256QAM. The results show that compared with the other three schemes, our proposed scheme has a better identification accuracy at low OSNR. In addition, the identification accuracy of this scheme can reach 100% when the OSNR ≥ 10 dB.

Reservoirs and the lateral seal of stratigraphic traps are controlled by the depositional environment or diagenesis. The recognition of facies and lithology from seismic attributes is an effective approach for identifying stratigraphic traps related to the depositional environment. In this paper, the occurrence of stratigraphic traps related to depositional environment in Permian aeolian clastics and Jurassic carbonate-evaporites was studied. To identify these stratigraphic traps, multiple seismic attributes were classified using supervised and unsupervised artificial neural networks (ANNs), which allowed the recognition of seismic facies and lithology. Neural networks are a powerful classification technique, which incorporates multiple attributes into a number of classes to identify sedimentary facies. Two algorithms comprising supervised and unsupervised neural networks are commonly implemented. With a supervised learning algorithm, prior information such as typical facies at the control wells are required to train the multilayer perceptron (MLP) network. With an unsupervised algorithm, only seismic data is input to the neural network, and competitive-learning techniques are employed to classify or self-organize the data based on its internal characteristics. Without prior information, the output classes are not labeled with lithofacies. According to the availability of prior information, supervised and unsupervised learning were applied to recognize dune-playa and carbonate-evaporite combinations, respectively. To characterize the depositional environments, joint interpretation with a geological model is necessary for both supervised and unsupervised classification. Two major findings have been derived from this work. First, the learning technology based on ANNs is effective to recognize sedimentary facies. The microfacies and lithologies identified by both supervised and unsupervised ANNs are very consistent with the drilled wells. Second, the recognition of depositional facies and lithology can characterize the stratigraphic traps in the study areas. Lateral seal plays a key role in stratigraphic traps. Playa siltstone and tight lagoonal limestone constitute the lateral seal in dune-playa and carbonate-evaporite combinations, respectively.

Accurate modulation identification of the received signals is undoubtedly a central component in multiple-input multiple-output (MIMO) communication systems, facilitating the demodulation task. This study presents a flexible and semi-supervised deep learning-driven strategy for automatic modulation identification. To this end, the multiclass classification problem is treated as multiple binary discrimination problems to address modulation identification challenges. Here, we merge the features extraction ability of the Generative Adversarial Network (GAN) model and the semi-supervised anomaly detection scheme, the one-class Support Vector Machine (1SVM). Essentially, a single GAN-based 1SVM detector is trained using training data of each class, with the samples of that class as inlier and all other samples as anomalies (i.e., one-vs.-rest). The 1SVM is trained using the features learned by the GAN model. A dataset consisting of three digital modulations (i.e., BFSK, CPFSK, and PAM4) and three analog modulations (i.e., AM-DSB, AM-SSB, and WB-FM), widely used in wireless communications systems, is employed to demonstrate the performance of the considered deep learning-based methods. Compared to Restricted Boltzmann Machine (RBM) and Deep Belief Network (DBN)-based 1SVM, the conventional GAN, DBN, and RBM with softmax layer as discriminator layer, the proposed GAN-based 1SVM detector offers superior discrimination performance of modulation types by achieving an averaged accuracy of 0.951 and F1-Score of 0.954. Results also showed that the GAN-1SVM detector dominates the state-of-the-art modulation classification techniques.

This paper presents a novel joint optical performance monitoring (OPM) scheme for multiple spatial modes in mode division multiplexing elastic optical network (MDM-EON) systems, which integrates the optimal Radon transform (ORT) image with the improved multi-task multi-input multi-output residual neural network (IMT-MIMO-ResNet) model. The ORT images effectively enhance the differentiation among various parameters across different spatial modes. Additionally, based on the output characteristics and intrinsic correlations among spatial modes of the MDM-EON system, we design the IMT-MIMO-ResNet model. The proposed scheme overcomes the limitations of existing approaches, where a single OPM device can monitor only one spatial mode. It enables simultaneous monitoring of bit rates (BRs), modulation formats (MFs), mode coupling (MC) intensities, and optical signal-to-noise ratio (OSNR) values for each MF across five spatial modes. Moreover, the scheme demonstrates favorable robustness under conditions with large chromatic dispersion, large differential group delay, and strong fiber nonlinear effects. Both simulation and experimental results show that the accuracy of baud rate identification (BRI) and modulation format identification (MFI) for each spatial mode reaches 100%, and the average mean absolute errors (MAEs) for OSNR values and MC intensities corresponding to each MF are below 0.4 dB and 0.010, respectively. The computational complexity of the proposed scheme, evaluated using floating-point operations (FLOPs) and parameters (Params), is approximately 1/7 of that required by a typical OPM scheme.

To understand the variants of Restricted Boltzmann Machines (RBMs) and Auto-encoders, we first need to outline what they are: RBMs are generative and unlike auto-encoders that are biased towards a limited set of data. RBMs are capable of generating new data with the joint set distribution. They are used to solve problems such as pattern recognition where there is a handwritten text that needs to be deciphered or a random pattern. It can also be used for recommendation engines where in collaboration with filtering techniques, recommendations are made to an end-user, and radar target recognition where it is used to detect intra-pulse with extremely low SNR and high noise. On the contrary, auto-encoders are not very commonly used in real-life applications; however, they are useful in reducing dimensionality and variational auto-encoders (VAE), where VAE learns the limitations of a probability distribution modeling the input data rather than learning the absolute function. RBMs and auto-encoders can be used for cyber-physical systems (CPS); RBMs are a dual-layer, two-part, erratic graphical model, allowing data to flow in two ways rather than one which forms the foundation of DBNs. CPS aims to use RBMs in making the model understand various functions, which will ultimately help to identify the hidden state and minimize the energy of the system. RBMs assign probabilities rather than definite values. Auto-encoders are unsupervised neural networks that use input vectors and try to match them to similar output vectors. These vectors are extremely skilled as they study compressed data encoding autonomously. This chapter will discuss in detail the breakthroughs with CPS and their findings. Previously CPS were only evaluated with techniques that did not differentiate the facts from the internal view, which ultimately resulted in a mismatch between the behavior of theoretical models and their real-life counterparts. This ultimately gave way to the question of how they could perform critical safety tasks. Another vital breakthrough related to CPS revolves around Intellectual Merit and Broader Impacts. They both revolve around the design and analysis of Artificial Intelligence as an integral part of CPS; the findings for these breakthroughs will be discussed in detail in the chapter.

Modulation format identification (MFI) and optical signal-to-noise ratio (OSNR) monitoring are essential for elastic optical networks. A method for joint modulation format identification and OSNR monitoring based on ternary neural networks was proposed. Further, a ternary neural network was established, and the constellation images after constant modulus algorithm (CMA) equalization were used as input features. Four commonly used modulation formats were distinguished, including dual-polarization (DP)-QPSK, 8QAM, 16QAM, and 64QAM. A 32G baud simulation system was constructed, and the results were analyzed. First, we investigated the influence of resolution and sample length on the identification accuracy. The values 32 and 7000 were selected for the two parameters, respectively, based on the balance between resource consumption and accuracy. Then, we compared the performance of the binary neural network (BNN), ternary neural network (TNN), and full-precision neural network (FNN). The results indicate that the memory consumption and extraction time of the TNN are similar to those of the BNN, and the accuracy is further improved. Moreover, the robustness of the system was analyzed. The results validate that the proposed method can tolerate fiber nonlinearity (from 500 km to 1500 km). Finally, an experiment was conducted to prove the practicality of the proposed method.

A lightweight two-stage convolutional (deep) neural network (CNN) based modulation format identification (MFI) scheme is proposed and demonstrated for the polarization domain multiplexing (PDM) fiber communication system with probabilistically shaped (PS) modulation formats. The scheme is tested on a PDM system at a symbol rate of 28 GBaud. Six probabilistically shaped (PS) modulation formats (of 3 bit/symbol PS-16QAM, PS-32QAM, and PS-64QAM, of 4 bit/symbol PS-32QAM and PS-64QAM, and of 5 bit/symbol PS-64QAM) along with six standard modulation formats (BPSK, QPSK, 8PSK and three uniformly shaped (US) QAM: US-16QAM, US-32QAM and US-64QAM) are identified by the trained CNN. By taking advantage of computer vision, the results show that the proposed scheme can provide very high accuracy and significantly improve the identification performance over the existing techniques. The influences of the learning rate of the CNN are also discussed.

Achievable information rates of optical communication systems are inherently limited by nonlinear distortions due to the Kerr effect occurred in optical fibres. These nonlinear impairments become more significant for communication systems with larger transmission bandwidths, closer channel spacing and higher-order modulation formats. In this paper, the efficacy of nonlinearity compensation techniques, including both digital back-propagation and optical phase conjugation, for enhancing achievable information rates in lumped EDFA- and distributed Raman-amplified fully-loaded C -band systems is investigated considering practical transceiver limitations. The performance of multiple modulation formats, such as dual-polarisation quadrature phase shift keying (DP-QPSK), dual-polarisation 16 -ary quadrature amplitude modulation (DP-16QAM), DP-64QAM and DP-256QAM, has been studied in C -band systems with different transmission distances. It is found that the capabilities of both nonlinearity compensation techniques for enhancing achievable information rates strongly depend on signal modulation formats as well as target transmission distances.

Artificial Neural Networks (ANN) is becoming increasingly important in the medical field for diagnostic applications. The popularity of ANN is mainly due to the high accuracy and the nominal convergence rate. But, the major drawback is that these characteristic features are not simultaneously available in the same network. While supervised neural networks are highly accurate, the requirement for convergence time is high. On the other hand, unsupervised neural networks are sufficiently faster but less accurate. This problem is tackled in this work by proposing a Modified Neural Network (MNN) which possesses the features of both the supervised neural network and the unsupervised neural network. The applicability of this network is explored in the context of abnormal retinal image classification. Images from four abnormal categories such as Non-Proliferative Diabetic Retinopathy (NPDR), Choroidal Neo-Vascularization Membrane (CNVM), Central Serous Retinopathy (CSR) and Central Retinal Vein Occlusion (CRVO) are used in this work. Suitable features are extracted from these images and further used for the training and testing process of the proposed ANN. Experimental results are analyzed in terms of classification accuracy and convergence rate. The experimental results are also compared with the results of conventional networks such as Back Propagation Network (BPN) and the Kohonen Network (KN). The results of the proposed modified ANN are promising in terms of the performance measures.

To adapt to the network dynamics and support elastic network resource allocation in the future elastic optical networks, optical transmitter with flexibility and re-configurability in modulation formats becomes indispensable, offering network providers flexibility in transport bandwidth. In this paper, we discuss our proposed two flexible high-order quadrature amplitude modulation (QAM) transmitter schemes, including (i) a flexible transmitter based on tandem in-phase/quadrature (IQ) modulators for generating 16QAM, 32QAM, 64QAM signals; and (ii) a scheme based on monolithically-integrated quad-Mach-Zehnder IQ modulator (QMZ-IQ) with binary driving electronics for synthesizing minimum phase-shift keying (MSK), quadrature phase-shift-keying (QPSK), 8-ary phase-shift keying (8PSK), and 16QAM signals. These schemes provide different approaches to configure a flexible high-order QAM transmitter, either using cascaded off-the-shelf modulators, or via highly-integrated monolithic modulator. These flexible and reconfigurable transmitter schemes provide the network providers different flexible transmitter solutions suitable for different application scenarios in elastic optical networks.

As human society to the information in the process of moving and growing demand for bandwidth communications capacity, the optical of new modulation formats increasingly attention and quickly play an important role in optical communications. How can the system bit error rate within a certain degree of stability while still maintaining high-speed long-distance dispersal system, has been a popular issue is the optical communications industry. Starting from the optical modulation format herein, the generation process of the system introduced various optical signal modulation format, the optical signal through the optical fiber was studied and the performance of the simulation, on the basis of the design of advanced optical modulation formats in an optical fiber communication system .

There are many modulation formats (such as FSK, PSK, ASK) in the process of underwater acoustic communication, and the task of the signal modulation identification is to distinguish the format that used from signals which received from acoustic communication. This technique for communication management, such as underwater acoustic communication spectrum monitoring, interference listening and signal confirmation, is of great significance. The cyclic cumulant is served as the characteristics of the signals, and the signal modulation formats is classified by support vector machine. We use the method to carry out the simulation experiment in alpha stable distribution noise environment and the sea experiment. The results show that the method is reasonable and effective.

Transcranial temporal interfering stimulation (tTIS) can focally stimulate deep parts of the brain related to specific functions using beats at two high frequencies that do not individually affect the human brain. However, the complexity and nonlinearity of the simulation limit it in terms of calculation time and optimization precision. We propose a method to quickly optimize the interfering current value of high‐definition electrodes, which can finely stimulate the deep part of the brain, using an unsupervised neural network (USNN) for tTIS. We linked a network that generates the values of electrode currents to another network, which is constructed to compute the interference exposure, for optimization by comparing the generated stimulus with the target stimulus. Further, a computational study was conducted using 16 realistic head models. We also compared tTIS with transcranial alternating current stimulation (tACS), in terms of performance and characteristics. The proposed method generated the strongest stimulation at the target, even when targeting deep areas or performing multi‐target stimulation. The high‐definition tTISl was less affected than tACS by target depth, and mis‐stimulation was reduced compared with the case of using two‐pair inferential stimulation in deep region. The optimization of the electrode currents for the target stimulus could be performed in 3 min. Using the proposed USNN for tTIS, we demonstrated that the electrode currents of tTIS can be optimized quickly and accurately. Moreover, we confirmed the possibility of precisely stimulating the deep parts of the brain via transcranial electrical stimulation.

Nowadays, classical strategies for fault detection and isolation present the lack of availability in order to tackle unknown scenarios. In here, fault localization based upon unsupervised neural networks is presented as an alternative approach in order to determine abnormal situations under certain conditions. This proposal uses two well-known techniques in order to classify patterns and clusters of patterns. This approach is followed in order to enhance the capability of fault localization by avoiding noise ratio and time variant behaviour. Both cluster techniques are connected in cascade mode. These two techniques are trained offline in order to response to on-line time constraints based upon case study