Hidden Nodes Research Articles

Machine learning the deuteron

We use machine learning techniques to solve the nuclear two-body bound state problem, the deuteron. We use a minimal one-layer, feed-forward neural network to represent the deuteron S- and D-state wavefunction in momentum space, and solve the problem variationally using ready-made machine learning tools. We benchmark our results with exact diagonalisation solutions. We find that a network with 6 hidden nodes (or 24 parameters) can provide a faithful representation of the ground state wavefunction, with a binding energy that is within 0.1% of exact results. This exploratory proof-of-principle simulation may provide insight for future potential solutions of the nuclear many-body problem using variational artificial neural network techniques.

How to model European electricity load profiles using artificial neural networks

We present a method to create synthetic, weather-dependent, annual electricity load profiles for European countries in hourly resolution using artificial neural networks as a necessary basis for long-term forecasts. To this end, we train fully connected dense artificial neural networks with 5 hidden layers and 1,024 hidden nodes per layer using historic data for Germany from 2006 to 2015. Input parameters used in the model comprise calendrical information, annual peak loads and weather data. We benchmark our results against the current state-of-the-art method to generate synthetic load profiles used in mid-term adequacy forecasts published by the European Network of Transmission System Operators (entso-e). For validation year 2016, our approach shows a mean absolute percentage error of 2.8%, whereas the method as used by entso-e shows an average error of 4.8%. We then conduct forecasts for Germany, Sweden, Spain, and France using our synthetic load profiles for scenario year 2025 to demonstrate their pan-European applicability. Finally, we assess parameter variations that demonstrate high influences of outdoor temperatures and wind speed on the electricity load. Our approach can help to increase prediction accuracy of future electricity loads as electricity load profiles are a necessary input for these forecasts.

Modeling and simulation of the IEEE 802.11e wireless protocol with hidden nodes using Colored Petri Nets

Wireless technologies are continuously evolving, including features such as the extension to mid- and long-range communications and the support of an increasing number of devices. However, longer ranges increase the probability of suffering from hidden terminal issues. In the particular case of Wireless Local Area Networks (WLANs), the use of Quality of Service (QoS) mechanisms introduced in IEEE 802.11e compromises scalability, exacerbates the hidden node problem, and creates congestion as the number of users and the variety of services in the network grow. In this context, this paper presents a configurable Colored Petri Net (CPN) model for the IEEE 802.11e protocol with the aim of analyzing the QoS support in mid- and long-range WLANs The CPN model covers the behavior of the protocol in the presence of hidden nodes to examine the performance of the RTS/CTS exchange in scenarios where the QoS differentiation may involve massive collision chains and high delays. Our CPN model sets the basis for further exploring the performance of the various mechanisms defined by the IEEE 802.11 standard. We then use this CPN model to provide a comprehensive study of the effectiveness of this protocol by using the simulation and monitoring capabilities of CPN Tools.

Development and validation of an artificial neural network prognostic model after gastrectomy for gastric carcinoma: An international multicenter cohort study.

BackgroundRecently, artificial neural network (ANN) methods have also been adopted to deal with the complex multidimensional nonlinear relationship between clinicopathologic variables and survival for patients with gastric cancer. Using a multinational cohort, this study aimed to develop and validate an ANN‐based survival prediction model for patients with gastric cancer.MethodsPatients with gastric cancer who underwent gastrectomy in a Chinese center, a Japanese center, and recorded in the Surveillance, Epidemiology, and End Results database, respectively, were included in this study. Multilayer perceptron neural network was used to develop the prediction model. Time‐dependent receiver operating characteristic (ROC) curves, area under the curves (AUCs), and decision curve analysis (DCA) were used to compare the ANN model with previous prediction models.ResultsAn ANN model with nine input nodes, nine hidden nodes, and two output nodes was constructed. These three cohort's data showed that the AUC of the model was 0.795, 0.836, and 0.850 for 5‐year survival prediction, respectively. In the calibration curve analysis, the ANN‐predicted survival had a high consistency with the actual survival. Comparison of the DCA and time‐dependent ROC between the ANN model and previous prediction models showed that the ANN model had good and stable prediction capability compared to the previous models in all cohorts.ConclusionsThe ANN model has significantly better discriminative capability and allows an individualized survival prediction. This model has good versatility in Eastern and Western data and has high clinical application value.

A Novel Method of Curve Fitting Based on Optimized Extreme Learning Machine

ABSTRACT In this article, we present a new method based on extreme learning machine (ELM) algorithm for solving nonlinear curve fitting problems. Curve fitting is a computational problem in which we seek an underlying target function with a set of data points given. We proposed that the unknown target function is realized by an ELM with introducing an additional linear neuron to correct the localized behavior caused by Gaussian type neurons. The number of hidden layer neurons of ELM is a crucial factor to achieve a good performance. An evolutionary computation algorithm–particle swarm optimization (PSO) technique is applied to determine the optimal number of hidden nodes. Several numerical experiments with benchmark datasets, simulated spectral data and measured data from high energy physics experiments have been conducted to test the proposed method. Accurate fitting has been accomplished for various tough curve fitting tasks. Comparing with the results of other methods, the proposed method outperforms the traditional numerical-based technique. This work clearly demonstrates that the classical numerical analysis problem-curve fitting can be satisfactorily resolved via the approach of artificial intelligence.

Smoothing L0 Regularization for Extreme Learning Machine

Extreme learning machine (ELM) has been put forward for single hidden layer feedforward networks. Because of its powerful modeling ability and it needs less human intervention, the ELM algorithm has been used widely in both regression and classification experiments. However, in order to achieve required accuracy, it needs many more hidden nodes than is typically needed by the conventional neural networks. This paper considers a new efficient learning algorithm for ELM with smoothing L0 regularization. A novel algorithm updates weights in the direction along which the overall square error is reduced the most and then this new algorithm can sparse network structure very efficiently. The numerical experiments show that the ELM algorithm with smoothing L0 regularization has less hidden nodes but better generalization performance than original ELM and ELM with L1 regularization algorithms.

Performance Analysis of Full Duplex MAC protocols for Wireless Local Area Networks with Hidden Node Collisions

Performance Analysis of Full Duplex MAC protocols for Wireless Local Area Networks with Hidden Node Collisions

Introducing Hidden Nodes in a Relationship Graph of Instagram Users

Introducing Hidden Nodes in a Relationship Graph of Instagram Users

Design of Extreme Learning Machine with Smoothed ℓ0 Regularization

In extreme learning machine (ELM), a large number of hidden nodes are required due to the randomly generated hidden layer. To improve network compactness, the ELM with smoothed l0 regularizer (ELM-SL0 for short) is studied in this paper. Firstly, the l0 regularization penalty term is introduced into the conventional error function, such that the unimportant output weights are gradually forced to zeros. Secondly, the batch gradient method and the smoothed l0 regularizer are combined for training and pruning ELM. Furthermore, both the weak convergence and strong convergence of ELM-SL0 are investigated. Compared with other existing ELMs, the proposed algorithm obtains better performance in terms of estimation accuracy and network sparsity.

Network reconstruction from binary-state time series in presence of time delay and hidden nodes

Complex networks with binary-state dynamics represent many meaningful behaviors in a variety of contexts. Reconstruction of networked systems hosting delayed binary processes with hidden nodes becomes an outstanding challenge in this field. To address this issue, we extend the statistical inference method to complex networked systems with distinct binary-state dynamics in presence of time delay and missing data. By exploiting the expectation-maximization (EM) algorithm, we implement the statistical inference based approach to different (i.e., random, small world, and scale-free) networks hosting delayed-binary processes. Our framework is completely data driven, and does not require any a prior knowledge about the detailed dynamical process on the network; especially, our method can independently infer each physical connectivity and estimate the time delay solely from the data of a pair of nodes in this link. We provide a physical understanding of the underlying mechanism; and extensive numerical simulations validate the robustness, efficiency, and accuracy of our method.

Acoustic Seabed Classification Based on Multibeam Echosounder Backscatter Data Using the PSO-BP-AdaBoost Algorithm: A Case Study From Jiaozhou Bay, China

When backpropagation neural network (BPNN) is often applied to supervised classification, problems arise, including a slow convergence rate, local extremum, and difficulty in determining the number of hidden layers and hidden nodes that affect the classification accuracy and efficiency. These problems can be overcome by using smarter network designs. Adaptive boosting (AdaBoost), which combines multiple weak classifiers to create a strong classifier, has a strong classification advantage. In this article, we propose an acoustic seabed classification method that combines AdaBoost with the particle swarm optimization (PSO). The PSO-BP-AdaBoost algorithm uses multibeam echosounder backscatter data to solve the multiclassification problem of diverse seafloor sediment types with small differences between types. We optimize a BPNN using the PSO algorithm to obtain the optimal initial weight and threshold and combine these to form an AdaBoost strong classifier. The input data is obtained from the sonar mosaic from multibeam echosounder backscatter data collected in Jiaozhou Bay using a series of fine processing techniques. These processing techniques result in 34-dimensional (34-D) features using ReliefF analysis. The most advantageous 8-D features are used as input into the AdaBoost algorithm based on one-level decision tree, PSO-BP algorithm, support vector machine (SVM), and PSO-BP-AdaBoost algorithm. The PSO-BP-AdaBoost classification model has better classification accuracy. The overall accuracy is improved by 12.68%, 6.78%, and 3.56%, respectively, which demonstrates that the PSO-BP-AdaBoost algorithm can be effectively applied to acoustic seabed classification and identification and achieves high precision.

Optimizing various parameters influencing mechanical properties of banana/coir natural fiber composites using grey relational analysis and artificial neural network models

The natural fibers of banana and coir from southern part of India was taken as the reinforcement material. Taguchi-based grey relational analysis was used for optimizing the factors influencing tensile, flexural and impact strength of banana/coir composites. Banana wt%, coir wt%, alkali treatment %, pressure and temperature during compression moulding process were the five major factors used for optimizing the mechanical strengths of epoxy-based composites. Grey relational analysis observed 20% banana/15% coir/3% alkali treatment/16 MPa pressure and 100°C temperature combination having the maximum mechanical properties in 16 trial combinations. The signal-to-noise ratio table of grey relational analysis showed combination with 20% banana /15% coir/5% alkali treatment/16 MPa pressure and 100°C temperature having optimized mechanical properties. Banana has the highest rank in improving the mechanical strength of natural composites, followed by coir, alkali treatment %, pressure and temperature. The graph between actual and artificial neural network computed results observed high correlation, proving that network topology of 5-3-1 with five factors, three hidden nodes and one response is the best method for analyzing the parameters in mechanical strength. Hybrid fiber composites with NaOH treatment showed hackles in interface with improved interfacial bonding during SEM analysis.

Cascading Attacks on Wi-Fi Networks: Theory and Experiments

We unveil the existence of a vulnerability in Wi-Fi (802.11) networks, which allows an adversary to remotely launch a denial-of-service (DoS) attack that propagates both in time and space. This vulnerability stems from a coupling effect induced by hidden nodes. Cascading DoS attacks can congest an entire network and do not require the adversary to violate any protocol. We demonstrate the feasibility of such attacks through experiments with real Wi-Fi cards, theoretical analysis, and ns-3 simulations. The experiment shows that an attacker can cause the throughput of a node outside its communication range to vanish. To gain insight into the root causes of the attack, we model the network as a dynamical system and analyze its limiting behavior and stability. The model predicts that a phase transition (and hence a cascading attack) is possible in linear networks when the retry limit parameter of Wi-Fi is greater or equal to 7, and also characterizes the phase transition region in terms of the system parameters.

Artificial neural networks applied to quality-by-design: From formulation development to clinical outcome

Quality-by-Design (QbD) is a methodology used to build quality into products and is characterized by a well-defined roadmap. In this study, the application of Artificial Neural Networks (ANNs) in the QbD-based development of a test drug product is presented, where material specifications are defined and correlated with its performance in vivo. Along with other process parameters, drug particle size distribution (PSD) was identified as a critical material attribute and a three-tier specification was needed. An ANN was built with only five hidden nodes in one hidden layer, using hyperbolic tangent functions, and was validated using a random holdback of 33% of the dataset. The model led to significant and valid prediction formulas for the three responses, with R2 values higher than 0.94 for all responses, both for the training and the validation datasets. The prediction formulas were applied to contour plots and tight limits were set based on the design space and feasible working area for the drug PSD, as well as for process parameters. The manufacturing process was validated through the production of three exhibit batches of 180,000 tablets in the industrial GMP facility, and the ANN model was applied to successfully predict the in vitro dissolution, with a bias of approximately 5%. The product was then tested on two clinical studies (under fasting and fed conditions) and the criteria to demonstrate bioequivalence to the Reference Listed Drug were met. In this study, ANNs were successfully applied to support the establishment of drug specifications and limits for process parameters, bridging the formulation development with in vitro performance and the positive clinical results obtained in the bioequivalence studies.

High precision variational Bayesian inference of sparse linear networks

Sparse networks can be found in a wide range of applications, such as biological and communication networks. Inference of such networks from data has been receiving considerable attention lately, mainly driven by the need to understand and control internal working mechanisms. However, while most available methods have been successful at predicting many correct links, they also tend to infer many incorrect links. Precision is the ratio between the number of correctly inferred links and all inferred links, and should ideally be close to 100%. For example, 50% precision means that half of inferred links are incorrect, and there is only a 50% chance of picking a correct one. In contrast, this paper infers links of discrete-time linear networks with very high precision, based on variational Bayesian inference and Gaussian processes. Our method can handle limited datasets, does not require full-state measurements and effectively promotes both system stability and network sparsity. On several of examples, Monte Carlo simulations illustrate that our method consistently has 100% or nearly 100% precision, even in the presence of noise and hidden nodes, outperforming several state-of-the-art methods. The method should be applicable to a wide range of network inference contexts, including biological networks and power systems.

Measuring Traffic Volumes Using an Autoencoder with No Need to Tag Images with Labels

Almost all vision technologies that are used to measure traffic volume use a two-step procedure that involves tracking and detecting. Object detection algorithms such as YOLO and Fast-RCNN have been successfully applied to detecting vehicles. The tracking of vehicles requires an additional algorithm that can trace the vehicles that appear in a previous video frame to their appearance in a subsequent frame. This two-step algorithm prevails in the field but requires substantial computation resources for training, testing, and evaluation. The present study devised a simpler algorithm based on an autoencoder that requires no labeled data for training. An autoencoder was trained on the pixel intensities of a virtual line placed on images in an unsupervised manner. The last hidden node of the former encoding portion of the autoencoder generates a scalar signal that can be used to judge whether a vehicle is passing. A cycle-consistent generative adversarial network (CycleGAN) was used to transform an original input photo of complex vehicle images and backgrounds into a simple illustration input image that enhances the performance of the autoencoder in judging the presence of a vehicle. The proposed model is much lighter and faster than a YOLO-based model, and accuracy of the proposed model is equivalent to, or better than, a YOLO-based model. In measuring traffic volumes, the proposed approach turned out to be robust in terms of both accuracy and efficiency.

Downsizing and enhancing broad learning systems by feature augmentation and residuals boosting

Recently, a broad learning system (BLS) has been theoretically and experimentally confirmed to be an efficient incremental learning system. To get rid of deep architecture, BLS shares the same architecture and learning mechanism of the well-known functional link neural networks (FLNN), but works in broad learning way on both the randomly mapped features of original features of data and their randomly generated enhancement nodes. As such, BLS often requires a huge heap of hidden nodes to achieve the prescribed or satisfactory performance, which may inevitably cause both overwhelming storage requirement and overfitting phenomenon. In this study, a stacked architecture of broad learning systems called D&BLS is proposed to achieve enhanced performance and simultaneously downsize the system architecture. By boosting the residuals between previous and current layers and simultaneously augmenting the original input space with the outputs of the previous layer as the inputs of current layer, D&BLS stacks several lightweight BLS sub-systems to guarantee stronger feature representation capability and better classification/regression performance. Three fast incremental learning algorithms of D&BLS are also developed, without the need for the whole re-training. Experimental results on some popular datasets demonstrate the effectiveness of D&BLS in the sense of both enhanced performance and reduced system architecture.

Efficient routing in UASN during the thermohaline environment condition to improve the propagation delay and throughput

In underwater acoustic sensor network (UASN), the challenging issues are bandwidth, higher propagation delay and heavy packet loss during data transmission. The issues can be solved through efficient routing algorithms. The existing UASN routing algorithms have larger latency in the network link and high rate of packet loss because of the salinity and temperature in the water at different depths. The salinity and temperature changes according to the depth and called as thermohaline circulation. In this paper, convex directional flooding optimisation (CDFO) algorithm improves the latency, throughput and lifetime of the nodes in the network under thermohaline condition and longshore drift from longshore current, which consist of transportation of sediments. The CDFO combines the convex optimisation and directional flooding-based routing algorithm, convex optimisation helps in identification of the hidden nodes in the network and strong communication links are established through polynomial time and semantic analysis and directional flooding algorithm reduces the packet loss and increases the network throughput. The routing protocol has implemented in ns2-AquaSim simulator and test bed for measurement of the performance metrics of the UASN.

RPCP‐MAC: Receiver preambling with channel polling MAC protocol for underwater wireless sensor networks

SummaryTo design a reliable and energy efficient medium access control (MAC) protocol for underwater wireless sensor networks (UWSNs) is an active research area due to its variety of applications. There are many issues associated with underwater acoustic channels including long and variable propagation delay, attenuation, and limited bandwidth which pose significant challenges in the design of MAC protocol. The available sender‐initiated asynchronous preamble‐based MAC protocols for UWSNs are not reliable and energy‐efficient. This is due to the problems caused by transmission of preambles for longer duration and collision of preambles from hidden nodes in sender‐initiated preamble‐based MAC protocols. To resolve these issues, the paper proposed an asynchronous receiver‐initiated preamble‐based MAC protocol named Receiver Preambling with Channel Polling MAC (RPCP‐MAC) protocol for shallow underwater monitoring applications with high data rates. The protocol is proposed to resolve data packet collision and support reliability in an energy‐efficient way without using any transmission schedule. The proposed protocol is based on the following mechanisms. Firstly, receiver preambling mechanism is adopted to reduce idle listening. Secondly, channel polling mechanism is used to determine missing data frame during its sleeping period and to minimize the active time of node and reduces energy wastage. Finally, a back‐off mechanism is applied to resolve collision when preambles are received simultaneously. In addition, performance analysis through Markov chain together with its validation with simulation‐based studies is reported in the paper. Both the analytical and simulation results have demonstrated the reliability achievable with RPCP‐MAC while providing good energy efficiency.

Event-Driven Data Gathering in Pure Asynchronous Multi-Hop Underwater Acoustic Sensor Networks.

In underwater acoustic modem design, pure asynchrony can contribute to improved wake-up coordination, thus avoiding energy-inefficient synchronization mechanisms. Nodes are designed with a pre-receptor and an acoustically adapted Radio Frequency Identification system, which wakes up the node when it receives an external tone. The facts that no synchronism protocol is necessary and that the time between waking up and packet reception is narrow make pure asynchronism highly efficient for energy saving. However, handshaking in the Medium Control Access layer must be adapted to maintain the premise of pure asynchronism. This paper explores different models to carry out this type of adaptation, comparing them via simulation in ns-3. Moreover, because energy saving is highly important to data gathering driven by underwater vehicles, where nodes can spend long periods without connection, this paper is focused on multi-hop topologies. When a vehicle appears in a 3D scenario, it is expected to gather as much information as possible in the minimum amount of time. Vehicle appearance is the event that triggers the gathering process, not only from the nearest nodes but from every node in the 3D volume. Therefore, this paper assumes, as a requirement, a topology of at least three hops. The results show that classic handshaking will perform better than tone reservation because hidden nodes annulate the positive effect of channel reservation. However, in highly dense networks, a combination model with polling will shorten the gathering time.