Multi-layer Optical Networks Research Articles

Cascadable optical nonlinear activation function based on Ge-Si.

To augment the capabilities of optical computing, specialized nonlinear devices as optical activation functions are crucial for enhancing the complexity of optical neural networks. However, existing optical nonlinear activation function devices often encounter challenges in preparation, compatibility, and multi-layer cascading. Here, we propose a cascadable optical nonlinear activation function architecture based on Ge-Si structured devices. Leveraging dual-source modulation, this architecture achieves cascading and wavelength switching by compensating for loss. Experimental comparisons with traditional Ge-Si devices validate the cascading capability of the new architecture. We first verified the versatility of this activation function in a MNIST task, and then in a multi-layer optical dense neural network designed for complex gesture recognition classification, the proposed architecture improves accuracy by an average of 23% compared to a linear network and 15% compared to a network with a traditional activation function architecture. With its advantages of cascadability and high compatibility, this work underscores the potential of all-optical activation functions for large-scale optical neural network scaling and complex task handling.

Nonlinear processing with linear optics

AbstractDeep neural networks have achieved remarkable breakthroughs by leveraging multiple layers of data processing to extract hidden representations, albeit at the cost of large electronic computing power. To enhance energy efficiency and speed, the optical implementation of neural networks aims to harness the advantages of optical bandwidth and the energy efficiency of optical interconnections. In the absence of low-power optical nonlinearities, the challenge in the implementation of multilayer optical networks lies in realizing multiple optical layers without resorting to electronic components. Here we present a novel framework that uses multiple scattering, and which is capable of synthesizing programmable linear and nonlinear transformations concurrently at low optical power by leveraging the nonlinear relationship between the scattering potential, represented by data, and the scattered field. Theoretical and experimental investigations show that repeating the data by multiple scattering enables nonlinear optical computing with low-power continuous-wave light. Moreover, we empirically find that scaling of this optical framework follows a power law.

Pre-sensor computing with compact multilayer optical neural network.

Moving computation units closer to sensors is becoming a promising approach to addressing bottlenecks in computing speed, power consumption, and data storage. Pre-sensor computing with optical neural networks (ONNs) allows extensive processing. However, the lack of nonlinear activation and dependence on laser input limits the computational capacity, practicality, and scalability. A compact and passive multilayer ONN (MONN) is proposed, which has two convolution layers and an inserted nonlinear layer, performing pre-sensor computations with designed passive masks and a quantum dot film for incoherent light. MONN has an optical length as short as 5 millimeters, two orders of magnitude smaller than state-of-the-art lens-based ONNs. MONN outperforms linear single-layer ONN across various vision tasks, off-loading up to 95% of computationally expensive operations into optics from electronics. Motivated by MONN, a paradigm is emerging for mobile vision, fulfilling the demands for practicality, miniaturization, and low power consumption.

Lensless opto-electronic neural network with quantum dot nonlinear activation

With the swift advancement of neural networks and their expanding applications in many fields, optical neural networks have gradually become a feasible alternative to electrical neural networks due to their parallelism, high speed, low latency, and power consumption. Nonetheless, optical nonlinearity is hard to realize in free-space optics, which restricts the potential of the architecture. To harness the benefits of optical parallelism while ensuring compatibility with natural light scenes, it becomes essential to implement two-dimensional spatial nonlinearity within an incoherent light environment. Here, we demonstrate a lensless opto-electrical neural network that incorporates optical nonlinearity, capable of performing convolution calculations and achieving nonlinear activation via a quantum dot film, all without an external power supply. Through simulation and experiments, the proposed nonlinear system can enhance the accuracy of image classification tasks, yielding a maximum improvement of 5.88% over linear models. The scheme shows a facile implementation of passive incoherent two-dimensional nonlinearities, paving the way for the applications of multilayer incoherent optical neural networks in the future.

Reinforcement-learning-based path planning in multilayer elastic optical networks [Invited

This paper reports our study on the multilayer path (MLP) planning method in multilayer networks to achieve the flexible accommodation of large-capacity and diversified traffic. In addition to traffic grooming of sub-lambda paths, MLP design requires optimal selection of the operational mode. In this paper, we discuss MLP design methods that use reinforcement learning and auxiliary graphs to achieve MLP designs that satisfy various requirements such as cost, energy, and low blocking probability. We introduce a heuristic MLP planning method using auxiliary graphs. This method can determine link weights of auxiliary graphs to comply with arbitrary predefined policies; it yields MLPs whose characteristics satisfy the MLP requirements. We then describe an approach to optimize the weights of the auxiliary graph using reinforcement learning. In simulations, we evaluate the number of successfully allocated MLP paths, the required number of transceivers, and the capacity distribution of optical paths in a scenario where MLP requests are generated sequentially. The quantitative results show that our MLP design method can adaptively adjust the link weights of the auxiliary graphs under various network conditions. This can significantly improve the performance compared to heuristic design methods that assume fixed policies.

Approximately Lossless Model Compression-Based Multilayer Virtual Network Embedding for Edge–Cloud Collaborative Services

Edge-cloud collaboration integrated with network virtualization is indispensable for diversified edge services. Meanwhile, the multilayer elastic optical network (ML-EON) is a promising underlying network for virtual network requests (VNRs) customized for edge-cloud collaborative services. However, the joint allocation of computing resources and high-dimensional ML-EON resources in virtual network embedding (VNE) will pose great computational complexity for online service deployment. In this paper, we propose an approximately lossless model compression mechanism to ease the computing burden of the VNE over ML-EON for edge-cloud collaborative services. An integer quadratic constraint programming (IQCP) model is established for the problem. Model compression based on Virtual Link Mapping Cost Estimation (VLMCE) is investigated to shield the variables and constraints related to ML-EON. In particular, the resource metric and topology metric are introduced into VLMCE to cope with resource contentions among virtual links in the same VNR and improve estimation accuracy. The model solving relies on Hopfield neural network (HNN) is further studied, where optimizing the compressed model is losslessly converted to minimizing the energy function of HNN. The experimental results reveal that the proposed mechanism guarantees an approximately lossless algorithm performance and a high time efficiency compared with the original IQCP model. The performances of VNR cost and blocking ratio are also promoted compared with the benchmarks.

Image sensing with multilayer nonlinear optical neural networks

Optical imaging is commonly used for both scientific and technological applications across industry and academia. In image sensing, a measurement, such as of an object's position, is performed by computational analysis of a digitized image. An emerging image-sensing paradigm breaks this delineation between data collection and analysis by designing optical components to perform not imaging, but encoding. By optically encoding images into a compressed, low-dimensional latent space suitable for efficient post-analysis, these image sensors can operate with fewer pixels and fewer photons, allowing higher-throughput, lower-latency operation. Optical neural networks (ONNs) offer a platform for processing data in the analog, optical domain. ONN-based sensors have however been limited to linear processing, but nonlinearity is a prerequisite for depth, and multilayer NNs significantly outperform shallow NNs on many tasks. Here, we realize a multilayer ONN pre-processor for image sensing, using a commercial image intensifier as a parallel optoelectronic, optical-to-optical nonlinear activation function. We demonstrate that the nonlinear ONN pre-processor can achieve compression ratios of up to 800:1 while still enabling high accuracy across several representative computer-vision tasks, including machine-vision benchmarks, flow-cytometry image classification, and identification of objects in real scenes. In all cases we find that the ONN's nonlinearity and depth allowed it to outperform a purely linear ONN encoder. Although our experiments are specialized to ONN sensors for incoherent-light images, alternative ONN platforms should facilitate a range of ONN sensors. These ONN sensors may surpass conventional sensors by pre-processing optical information in spatial, temporal, and/or spectral dimensions, potentially with coherent and quantum qualities, all natively in the optical domain.

Partitionable High-Efficiency Multilayer Diffractive Optical Neural Network.

A partitionable adaptive multilayer diffractive optical neural network is constructed to address setup issues in multilayer diffractive optical neural network systems and the difficulty of flexibly changing the number of layers and input data size. When the diffractive devices are partitioned properly, a multilayer diffractive optical neural network can be constructed quickly and flexibly without readjusting the optical path, and the number of optical devices, which increases linearly with the number of network layers, can be avoided while preventing the energy loss during propagation where the beam energy decays exponentially with the number of layers. This architecture can be extended to construct distinct optical neural networks for different diffraction devices in various spectral bands. The accuracy values of 89.1% and 81.0% are experimentally evaluated for MNIST database and MNIST fashion database and show that the classification performance of the proposed optical neural network reaches state-of-the-art levels.

Employing FAMTAR and AHB to Achieve an Optical Resource-Efficient Multilayer IP-Over-EON SDN Network

With the global Internet traffic continuously growing, network operators face more and more challenges related to the management of their networks. Efficient utilization of the available network resources becomes crucial to maintain the desired Quality of Service level and control the upsurge in operational expenses. The combination of the Software-Defined Networking concept with the multi-layer network architecture can simplify the process of control and management of the network, its layers, and resources. In this paper, we propose a solution to enhance resource utilization in software-defined multi-layer optical networks. The proposed solution takes the advantages of the Software-Defined Networking, Flow-Aware Multi-Topology Adaptive Routing, and Automatic Hidden Bypasses mechanisms to ensure simultaneous, multi-path data transmissions in both IP and optical layers. The Software-Defined Networking controller manages both mentioned mechanisms, selects the best possible bypass, and allocates lightpaths to ensure that the optical spectral efficiency is optimal. The evaluation shows that the proposed solution for multi-layer software-defined network increases the overall network performance and resource utilization.

Multi-layer optical Fourier neural network based on the convolution theorem

To take full advantage of the application of neural networks to optical systems, we design an optical neural network based on the principle of free-space optical convolution. In this article, considering the need for a high-power light source to excite the nonlinearity of an optical material, we describe how to reduce the power consumption of the system by quantifying the output of each layer after the softmax operation as an 8-bit value and loading these values into amplitude-only spatial light modulators (SLMs). In addition, we describe how to load the matrix with positive and negative values in the amplitude-only SLM by utilizing Fourier properties of the odd-order square matrix. We apply our six-layer optical network to the classification of Mixed National Institute of Standards and Technology database (MNIST) and Fashion-MNIST and find that the accuracy reaches 92.51% and 80.67%, respectively. Finally, we consider the error analysis, power consumption, and response time of our framework.

TAPI-enabled SDN control for partially disaggregated multi-domain (OLS) and multi-layer (WDM over SDM) optical networks [Invited

Network operators are facing a critical issue on their optical transport networks to deploy 5G + and Internet of Things services. They need to address the capacity increase by a factor of 10, while keeping a similar cost per user. Over the past years, network operators have been working on the optical disaggregated approach with great interest for achieving the required efficiency and cost reduction. In particular, partially disaggregated optical networks make it possible to decouple the transponders from the transport system (known as an open line system) that are provided by different vendors. On the other hand, space division multiplexing (SDM) has been proposed as the key technology to overcome the capacity crunch that the optical standard single-mode fibers are facing to support the forecasted 10 × growth. Spatial core switching is gaining interest because it makes it possible to deploy SDM networks to bypass the overloaded wavelength division multiplexing (WDM) networks, by provisioning spatial media channels between WDM nodes. This paper presents, to the best of our knowledge, the first experimental demonstration of transport-application-programming-interface-enabled software defined networking control architecture for partially disaggregated multi-domain and multi-layer (WDM over SDM) optical networks.

On the suitability, requisites, and challenges of machine learning [Invited

The introduction of 5G, the increasing number of connected devices, and the exponential growth of services relying on connectivity are pressuring multilayer networks to improve their scaling, efficiency, and controlling capabilities. However, enhancing those features consistently results in a significant amount of complexity in operating the resources available across heterogeneous vendors and technology domains. Thus, multilayer networks should become more intelligent in order to be efficiently managed, maintained, and optimized. In this context, we are witnessing an increasing interest in the adoption of artificial intelligence (AI) and machine learning (ML) in the design and operation of multilayer optical transport networks. This paper provides a brief introduction to key concepts in AI/ML, highlighting the conditions under which the use of ML is justified, on the requisites to deploy a data-driven system, and on the challenges faced when moving toward a production environment. As far as possible, some key concepts are illustrated using two realistic use-cases applied to multilayer optical networks: cognitive service provisioning and quality of transmission estimation.

A survey on multi-layer IP and optical Software-Defined Networks

As Software-Defined Networks become more and more popular, they start to appear in various architectures. The most common utilization of SDN is to control the network in the electric layers, including OSI/ISO layers 2 and above. However, SDN can be used much further. Professionals and scientists noticed that combining management and control of many layers at the same time can provide benefits. At one point, SDN started to be envisioned for multi-layer networks.In this survey we present, compare and contrast solutions that utilize SDN in multi-layer network architectures. The main objective is to analyze evolving multi-layer network architectures and show how these solutions coupled with SDN contribute to make future networks simple, flexible and cost-effective.

Survivable automatic hidden bypasses in Software-Defined Networks

The paper focuses on the problem of effective and resilient network resources management under condition of increasing Internet traffic. Multilayer automatic optical networks are considered as a solution usually indicated as the most appropriate for the future Internet. In this paper, we propose a Survivable Automatic Hidden Bypasses approach to enhance resource utilization and reliability in multilayer optical networks. Our solution uses the Software-Defined Networking concept to automatically create or remove hidden bypasses which are not visible at the network layer. We propose a novel survivability metric which is applied during optical bypass creation in order to handle network failures in a proactive manner. Such an approach contrasts with the most common one where bypasses are considered as a mechanism to handle network failures. We combine the metric with different restoration schemas and investigate the proposed mechanism under three failure scenarios. The mechanism of hidden bypasses increases throughput and reduces transmission delays while novel proactive survivability mechanisms neutralize negative impact of network failures.

Data-Oriented Task Scheduling in Fixed- and Flexible-Grid Multilayer Inter-DC Optical Networks: A Comparison Study

It is known that in multilayer interdatacenter optical networks (ML-IDCONs), network services can generate numerous data-oriented tasks (DoTs) that require not only inter-datacenter (DC) data transfers but also data processing in the DCs. In this paper, we perform a comparison study on the scheduling of DoTs (i.e., setting up lightpaths on fiber links for inter-DC data transfers and scheduling DoT buffering/processing in DCs) in fixed- and flexible-grid ML-IDCONs. We propose a DoT scheduling algorithm that can work well for both types of networks. Specifically, for each DoT, the algorithm first tries to serve it using the residual bandwidth in the IP layer. This is achieved by expanding the time-expanded network approach and transforming the store-and-forward assisted DoT scheduling problem into a minimum-cost maximum-flow (MCMF) problem. Then, by solving the MCMF problem, we find the way to maximize the data transfer throughput and minimize the total dc storage usage simultaneously. Next, if the obtained data transfer throughput is not sufficient for the DoT, the algorithm tries to build lightpath segments for it based on the branch and bound scenario. Extensive simulations are conducted to evaluate the proposed algorithm's performance in fixed- and flexible-grid ML-IDCONs, and also compare it with three benchmarks. Simulation results indicate that for DoT scheduling, the flexible-grid ML-IDCON can outperform fixed-grid ones in terms of the blocking probability, energy consumption of transponders, and usage of dc storage, and our algorithm achieves lower blocking probability than the benchmarks with comparable or higher time-efficiency.

Guaranteeing Traffic Survivability and Latency Awareness in Multilayer Network Design

The massive adoption of real-time Internet services (e.g., online gaming or video streaming) is threatening the way in which operators and service providers design and operate their networks. New requirements, such as maximum latency, are becoming de facto indicators to measure network quality, complementing the classical requisites of fault tolerance to increase network availability. However, finding fault-tolerant designs that also limit the end-to-end latency of IP flows is challenging in multilayer IP-over-WDM networks, since traffic routed over IP links depends on the actual sequences of the traversed fibers and how they change in each failure state. Despite its practical importance, the research on joint optimization of end-to-end latencies and fault tolerance in multilayer optical networks is minimal. In this context, this paper presents a planning algorithm for multilayer IP-over- WDM (with OSPF-ECMP routing in the IP layer), considering both survivability and end-to-end latency awareness. Three versions of the algorithm are provided, assuming three different recovery mechanisms in the optical layer: 1 + 1 lightpath protection, unprotected lightpaths, and lightpath restoration. In all cases, as is customary, the optical layer is assumed to react first to the failures, and after that, the IP/OSPF layer adapts to the surviving topology. The aim of our algorithm is to design minimum cost networks that guarantee not only 100% survivability under some representative failure states, but also the maximum end-to-end latency for IP flows in the case of any failures. According to our results, a careful and holistic design may achieve both objectives simultaneously, without, in most cases, noticeable side effects in terms of cost and/or throughput. In its turn, we see that, when end-to-end latencies are not considered, designs tend to produce very high latencies in some flows and failure states.

Grating-assisted vertical couplers for signal routing in multilayer integrated optical networks

Grating-assisted vertical couplers, which behave as add-drop filters, are proposed for wavelength routing of the signal among the different layers of on-chip multilayer optical networks. The device implements a 2×2 wavelength router which can be assembled into higher-order three-dimensional matrices. In particular, simple design criteria are found through a rapid and efficient optimization approach based on the mode analysis and demonstrated by the Finite Difference Time Domain (FDTD) simulations. The proposed numerical method is valid either for in-plane or for vertical grating-assisted couplers and it requires negligible computational effort. Different configurations of grating-assisted vertical couplers are designed and their spectral behavior is analyzed by the FDTD. The proposed devices achieve low values of the crosstalk between the different ports (below −20dB) and of the input reflection (below −15dB).

Maximized reliability with minimal cross-layer cutset under arbitrary link failure probability in multilayer optical networks

We investigate the maximized cross-layer reliability under arbitrary link failure probability in multilayer optical networks. A concept of minimal cross-layer cutset is first defined and a reliability model with arbitrary physical link failure probability is built in the multilayer optical networks. In order to reduce the scale of cutset enumeration, we introduce two metrics to estimate cross-layer reliability, i.e., the minimum cross-layer node reliability and the minimum cross-layer edge reliability (MCER). Furthermore, we develop two linear programming (LP) models and two heuristic algorithms to maximize the cross-layer reliability of multilayer optical networks, i.e., the minimum shared-risk mapping algorithm and the least shared failure probability algorithm. Simulation results show that: (i) the cross-layer reliability of the two proposed algorithms is close to the LP solutions under logical networks with different sizes, which achieves better results in terms of additional resources utilization compared with the shortest path algorithm; (ii) less difference between the results of our proposed algorithms and the results of the shortest path algorithm is accompanied by a small standard deviation of failure probability distribution. Moreover, the superiority of our proposed algorithms becomes more remarkable with the increasing of the standard deviation.

A 3D multilayer optical network on chip based on mesh topology

Optical Network-on-Chip (ONoC) is becoming a promising solution for high performance on chip interconnection, which draws much attention from many researchers. ONoC combined with 3D integration technology can address some issues of two-dimensional ONoC such as long distance and limited scalability, which have been shown to be effective solutions for further promoting the performance of ONoC. However, the infeasibility of most existing routers with four or five ports poses a problem in 3D optical interconnect as seven-port optical routers are required in 3D networks. To solve this problem, in this paper, we propose a 3D multilayer optical network on chip (3D MONoC) based on Votex, a non-blocking optical router with seven ports. We describe the optical router and the 3D network in detail. The proposed router architecture not only realizes 3D interconnection and can be utilized in most 3D ONoC, but also can be beneficial in achieving smaller area, lower cost of ONoC. We compare Votex with the traditional $$7\times 7$$7×7 optical router based on crossbar, which indicated that Votex can save cost. Moreover, we make a comparison of 3D MONoC employing Votex against its 2D counterpart. Simulation results show that the performance including ETE delay and throughput of 3D MONoC can be improved.

Generic Wavelength-routed Optical Router (GWOR) based on grating-assisted vertical couplers for multilayer optical networks

A Generic Wavelength-routed Optical Router (GWOR) based on grating-assisted vertical couplers is proposed to be used as 4×4 routing matrix in multilayer optical networks. The device exploits, as basic building blocks, four vertical grating-assisted couplers made of three vertically stacked waveguides. The central waveguide is patterned with a periodic Bragg grating that guarantees the wavelength routing of the signal at the Bragg wavelength. The design and the analysis of the grating-assisted vertical couplers, performed by two different numerical methods, the Bidirectional Beam Propagation Method based on the Method of Lines (MoL-BBPM) and the Finite Difference Time Domain (FDTD) method, are reported. Moreover, the GWOR matrix is analyzed, with a very limited computational effort, by suitably composing the numerically calculated transmittances of the 2×2 elementary building blocks. The proposed GWOR matrix achieves low values of the insertion loss (i.e. maximum insertion loss IL=0.2dB) and crosstalk below −15dB.