Neutral Network Research Articles

High-resolution reconstruction and a-priori modeling of turbulent flames in the context of large eddy simulation using the convolutional neural network

In large eddy simulation (LES) of turbulent combustion, accurate modeling of the unresolved scalar flux and filtered reaction source terms is challenging. In the present work, a convolutional neutral network (CNN) was developed for the high-resolution reconstruction of the unfiltered progress variable, velocity and reaction rate based on the filtered quantities that are available from LES or by filtering the direct numerical simulation (DNS) data. The unclosed terms in the filtered progress variable transport equation were then modeled using the reconstructed quantities from the proposed CNN model and the approximate deconvolution method (ADM). Two DNS cases, i.e., case A and case B, with different Karlovitz numbers (Ka) were performed to assess the performance of the models a−priori. The unfiltered and filtered DNS results were first presented. It was found that the small-scale wrinkling structures of the flames and turbulence are largely filtered out, and the reaction zone is broadened by filtering. Then, the progress variable, velocity and reaction rate were reconstructed from the filtered DNS data. The results of reconstruction by ADM and the CNN model were compared with those from the DNS. It was found that the distributions of various quantities predicted by the CNN model agree well with those of the DNS. Finally, the unresolved scalar flux and reaction source terms in the filtered progress variable transport equation were modeled. The statistics of the modeled results were analyzed and it was shown that the CNN model performs better than ADM. Overall, the CNN model is promising for the data reconstruction and model development of turbulent combustion.

Ultra-High Modulus Hydrogels Mimicking Cartilage of the Human Body.

The human body is comprised of numerous types of cartilage with a range of high moduli, despite their high hydration. Owing to the limitations of cartilage tissue healing and biological grafting procedures, synthetic replacements have emerged but are limited by poorly matched moduli. While conventional hydrogels can achieve similar hydration to cartilage tissues, their moduli are substantially inferior. Herein, triple network (TN) hydrogels are prepared to synergistically leverage intra-network electrostatic repulsive and hydrophobic interactions, as well as inter-network electrostatic attractive interactions. They are comprised of an anionic 1st network, a neutral 2nd network (capable of hydrophobic associations), and a cationic 3rd network. Collectively, these interactions act synergistically as effective, yet dynamic crosslinks. By tuning the concentration of the cationic 3rd network, these TN hydrogels achieve high moduli of ≈1.5 to ≈3.5MPa without diminishing cartilage-like water contents (≈80%), strengths, or toughness values. This unprecedented combination of properties poises these TN hydrogels as cartilage substitutes in applications spanning articulating joints, intervertebral discs (IVDs), trachea, and temporomandibular joint disc (TMJ).

Predicting the long-term CO2 concentration in classrooms based on the BO–EMD–LSTM model

Predicting the carbon dioxide (CO2) concentration in classrooms is important for undertaking preventive measures before it could harm health and learning efficiency of students. Existing studies have primarily focused on short-term predictions, and it is challenging to make long-term predictions of classroom CO2 concentrations. In this study, a novel divide-and-conquer method namely BO–EMD–LSTM is proposed, which integrates Bayesian optimization (BO), empirical mode decomposition (EMD) and long-short term memory (LSTM). In this model, the EMD algorithm is first applied to decompose the complex CO2 sequence into several subsequences, and the BO algorithm is then employed to determine the hyperparameters of the LSTM models. Our results indicate that the BO–EMD–LSTM method could improve the accuracy of long-term CO2 concentration predictions. Compared with traditional methods, such as LSTM network, back propagation neutral network, and random forest algorithm, this method could reduce the mean absolute error by more than 55% when making predictions 30 min ahead. Such mean absolute error reduction exceeded 60% when the prediction time was 5–15 min. Moreover, when making long-term predictions, the R-squared of BO–EMD–LSTM remained above 95%; however, it dropped to less than 70% when applying other methods. When the error variation is within the acceptable range, merging subsequences with similar frequencies can effectively reduce time consumption and computational burden. The proposed BO–EMD–LSTM method provides an effective tool to predict long-term CO2 concentrations and could aid decision makers implement CO2 control measures.

Characteristics of wideband tympanometry in patients with Ménière's disease based on neutral network

Objective:To construct a prediction model for Ménière's disease based on neural network and evaluate its prediction ability. Methods:Sixty-four patients with Ménière's disease underwent gadolinium enhanced magnetic resonance imaging of inner ear which showed endolymphatic hydrops. Meanwhile, 40 healthy adults were enrolled as controls. The database of wideband tympanometry of patients and control subjects was analyzed, and the neural network model was established by MATLAB 2021a software. The prediction ability of the model was evaluated by accuracy, positive predictive value, negative predictive value, the Youden index, sensitivity, specificity, receiver operating characteristic curve and area under curve （AUC）. Results:A feedforward network model was built with a single hidden layer to predict Ménière's disease with wideband tympanometry. There were 104 features in the input layer, 13 neuron nodes in the hidden layer and 1 output neuron in the output layer. The accuracy of the model was 83.2%, the positive predictive value was 80.7%, the negative predictive value was 84.3%, the sensitivity was 76.5%, the specificity was 83.7%, the Youden index was 0.602, and the AUC was 0.855. Conclusion:Based on neural network, the prediction model of Ménière's disease with high accuracy was constructed according to the results of wideband tympanometry, which provided reference for the diagnose of Ménière's disease.

Detection of Cardiac Arrhythmia Using Multi-Perspective Convolutional Neutral Network for ECG Heartbeat Classification

The leading death cause all over the world is heart disease. The presence of arrhythmias has to be examined to detect heart disease in early stage. The abnormality in heart beat rhythm is known as Arrhythmia. The speed of heart beat can be detected by Arrhythmia, it may be too slow, too fast or irregular patterns of heart beat are considered as Arrhythmia and there are various types of Arrhythmia. The electrocardiogram (ECG) produces signals, classification of such signals is very crucial for knowing the irregularity in the patterns of heart beat. As detection of arrhythmia is a challenging task, there is a great demand for an automatic detection technique to identify abnormal signals produced by heart which cannot be done manually. Therefore this paper provides a method for detection of Cardiac Arrhythmia using Multi-Perspective Convolutional Neutral Network (MPCNN) for ECG Heartbeat Classification. Basing on Physionet's MIT-BIH Arrhythmia Dataset the signals of ECG arrhythmia can be categorized into five classes. Number of layers, filters and size of the filter are appropriate parameters which are heuristically optimized for operating swiftly for operation of MPCNN effectively. Compared with the ultra-modern methods as Quantum Neural Networks and Deep Convolutional Neural Networks, the proposed method results in efficient performance having high level accuracy of 96.46%, 98.1% and 96.2% are the F1 scores for SVP and PVC respectively. The effective detection of heartbeat rhythm and irregularities can be identified using this model.

Data assimilation for the burnup distribution applying the three-dimensional variational and artificial neutral network algorithm

In this paper, a data-assimilation method has been proposed and applied for the burnup distribution of PWR. The burnup distribution is significant for the safety and economy of the reactor, as it is essential for the fuel-reloading design and optimization. Due to the fact that the burnup distribution cannot be measured directly during the reactor operation, the numerical simulation is widely applied in reactor engineering. However, there definitely exist differences between the numerical simulation and the actual core due to some unavoidable factors, such as component manufacturing deviation, uneven flow distribution and so on. These differences would induce the errors to the simulation values of power distributions and hence to the burnup distributions. Therefore, a data-assimilation method for the burnup distribution has been proposed, aiming at reducing the burnup-distribution errors with application of the power-distribution measurements. In our research, the three-dimensional variational (3DVAR) algorithm was applied for burnup-distribution calibration and the artificial neutral network (ANN) algorithm was applied to establish the complex relations between burnup distribution and power distribution. As engineering verification, the proposed data-assimilation method has been applied to the CNP1000 PWR operated in China. The numerical results indicated that the burnup-distribution errors can be reduced notably, as the maximum value of relative errors for power distribution can be reduced from 9.53% to 5.11%.

Evaluation of Teachers' Innovation and Entrepreneurship Ability in Universities Based on Artificial Neural Networks

Based on iceberg theory and the questionnaire of competency's elements, hierarchical index system of evaluation of teachers' innovation and entrepreneurship competency in universities is established. Through researches, the authors think that analytic hierarchy process (AHP) is a more scientific and reasonable evaluation method whose rationality is checked by satisfactory consistency while the evaluation model of artificial neutral network doesn't consider weighting. If the samples are more than 30, the evaluation of neural network model of teachers' innovation and entrepreneurship competency can achieve the accurate results and satisfactory requirements. Since the method of artificial neutral network has advantages of strong operability, simple rules, and minor errors, it can greatly reduce the workload because it not only eliminates human subjectivity of evaluation and greatly simplifies the process of evaluation, but also improves working efficiency and provides a new way of thinking for evaluation of the teachers' innovation and entrepreneurship competency in universities.

Beads and Globules from Fires: Can They Be Differentiated through Metallurgical Analysis Based on Machine Learning Algorithms?

Electrical apparatuses are prone to faults, which generally causes fires. During such fires, the identification of resolidified copper beads on wires has a strong influence on the direction of the fire investigation. There are four kinds of resolidified beads formed on copper conductors that have been through the fire with and without voltage, namely, ‘cause’ beads (CB), ‘victim’ beads (VB), overload globules (OG), and fire melting globules (FG). First, to improve the identification’s objectivity and quantifiability, we used various morphologic parameters of crystals and porosities to express metallurgical microcharacteristics, such as Ar-G, As-G, An-G, Dm-G, R-G, FD-G, Fm-G, Ar-G, As-P, An-P, Dm-P, R-P, FD-P, Fm-P, P3-P, and Cu2O. Then, several machine learning classifiers were developed to predict the melted beads based on metallurgical morphologic parameters by using SVM, BP neutral network (BPNN), AdaBoost, bagging, and random forest (RF), respectively. Models were trained and tested based on the sample set, consisting of 560 samples which were collected from real room fires. ACC/F1 of the RF model were 0.894/0.805, respectively, which are superior to SVM, BPNN, AdaBoost, and bagging. For the RF classifier, the recall rates of CB, VB, OG, and FG were 92.5%, 67.5%, 100%, and 97.5%, respectively, indicating that RF has best potential to predict OG and FG. The variable importance was analyzed to distinguish key features, and the results revealed that Cu2O has highest impact on bead classification. We cannot find much promise with this method that uses multiple metallurgical and morphological parameters for distinguishing between CB and VB. It is confirmed that no machine learning classifiers combined with metallurgical analysis could do this work well in this paper. Thus, we strongly recommend that other evidence for investigation in the room fire should also be considered to cover the shortage of this kind.

Relative contribution of multi-source water recharge to riparian wetlands along the lower Yellow River

Rivers play a vital role in both the formation and maintenance of riparian wetland hydrology. However, few studies have focused on the response of water recharge of riparian wetlands to altered hydrological processes induced by water-sediment regulation practices. To fill this gap, our study investigated the contribution of multi-source water recharge of riparian wetlands in the lower Yellow River, as well as its influence both during and before the water-sediment regulation scheme of Xiaolangdi Dam. Our study is based on hydrochemistry and isotopic methods, using a Bayesian mixing model and artificial neutral network model. The results showed that riparian wetlands were fed by mixed sources, including groundwater, canals, the Yellow River, and precipitation. However, seasonal evaporation introduced additional variation, which affected the relative contribution of these sources across seasons. Among these sources, the Yellow River served as the main water source for recharging riparian wetlands, and its contribution varied both spatially and temporally (across seasons). Specifically, proximity of riparian wetlands was the primary factor explaining spatial variation in the contribution of Yellow River, while climatic (12.38%) and hydrological variabilities (87.62%) explained seasonal variation. Among these climatic and hydrological variables, suspended sediment content was the most important factor—with a relative contribution of 36.33%. By determining the contribution of the Yellow River to the recharge of riparian wetlands, our study has provided information which is beneficial to adaptive management of river-fed riparian wetlands, especially under the implementation of water-sediment regulation practices.

Experimental exploration of a ribozyme neutral network using evolutionary algorithm and deep learning

A neutral network connects all genotypes with equivalent phenotypes in a fitness landscape and plays an important role in the mutational robustness and evolvability of biomolecules. In contrast to earlier theoretical works, evidence of large neutral networks has been lacking in recent experimental studies of fitness landscapes. This suggests that evolution could be constrained globally. Here, we demonstrate that a deep learning-guided evolutionary algorithm can efficiently identify neutral genotypes within the sequence space of an RNA ligase ribozyme. Furthermore, we measure the activities of all 216 variants connecting two active ribozymes that differ by 16 mutations and analyze mutational interactions (epistasis) up to the 16th order. We discover an extensive network of neutral paths linking the two genotypes and reveal that these paths might be predicted using only information from lower-order interactions. Our experimental evaluation of over 120,000 ribozyme sequences provides important empirical evidence that neutral networks can increase the accessibility and predictability of the fitness landscape.

A generic framework for augmented concrete mix design: Optimisation of geopolymer concrete considering environmental, financial and mechanical properties

While concrete mix design has typically focused on the development of low-cost concretes that achieve a particular set of material characteristics, it is now becoming increasingly important to consider environmental impact. This is a challenging problem because concrete mix design, particularly when involving waste materials, is impacted not only by the concrete mix performance but also on material availability and local variation in chemical composition. This variation impacts material performance, financial cost and environmental sustainability. This paper presents a generic framework for mix design optimisation based on a material design and profiling tool coupled with an optimisation algorithm. The framework is then demonstrated for geopolymer concrete (GPC) using a new data-driven mix design tool based on an artificial neutral network to predict mechanical properties (compressive strength and slump) coupled with life-cycle assessment and financial models to quantify mix performance. Together these inputs are used as the basis for single and multi-objective optimisation using a genetic algorithm as part of set of scenarios examining the impact of regional scale transportation scenarios and the chemical composition of binders on the resultant mix properties. The framework's implementation for GPC enabled up to a 50% reduction in CO2-eq emissions compared to equivalent strength experimental mixes, and demonstrated that for a given strength, slump and transportation case the overall CO2-eq emissions can increase by up to 30% where multi-objective optimisation is used (financial and CO2-eq emissions minimization) when compared to minimization of CO2-eq emissions alone.

A hybrid intelligent genetic algorithm for truss optimization based on deep neutral network

The truss optimization problem has been extensively investigated, and the optimized trusses have been widely used in various fields. Truss optimization is a challenging optimization involving many difficulties, such as discrete variables and non-convex problems. In order to solve these problems, various metaheuristic optimization methods have been proposed. Due to the stochastic feature of these methods, it is always computationally intensive, and the optimization results may vary greatly in different optimization runs. In this paper, a hybrid intelligent genetic algorithm (HIGA) is proposed to improve the effectiveness and efficiency of truss optimization problems. This method systematically integrates deep neural network (DNN) and genetic algorithm (GA) in the optimization process. A two-step training procedure is proposed where the data generated during the optimization process is exploited to update DNN. Next, based on the generated DNN model, an optimization prediction procedure is proposed to seek more optimized trusses in an efficient manner. Through the investigation of three classical truss problems (size optimization, shape optimization and size-shape integrated optimization), the effectiveness of the proposed method is validated. In addition, the influence of different settings on the optimization performance and efficiency is investigated to demonstrate the applicability and robustness of the proposed method.

A Hybrid Machine Learning Model Coupling Double Exponential Smoothing and ELM to Predict Multi-Factor Landslide Displacement

The deformation of landslides is a non-linear dynamic and complex process due to the impacts of both inherent and external factors. Understanding the basis of landslide deformation is essential to prevent damage to properties and losses of life. To forecast the landslides displacement, a hybrid machine learning model is proposed, in which the Variational Modal Decomposition (VMD) is implemented to decompose the measured total surface displacement into the trend and periodic components. The Double Exponential Smoothing algorithm (DES) and Extreme Learning Machine (ELM) were adopted to predict the trend and the periodic displacement, respectively. Particle Swarm Optimization (PSO) algorithm was selected to obtain the optimal ELM model. The proposed method and implementation procedures were illustrated by a step-like landslide in the Three Gorges Reservoir area. For comparison, Least Square Support Vector Machine (LSSVM) and Convolutional Neutral Network–Gated Recurrent Unit (CNN–GRU) were also conducted with the same dataset to forecast the periodic component. The application results show that DES-PSO-ELM outperformed the other two methods in landslide displacement prediction, with RMSE, MAE, MAPE, and R2 values of 1.295mm, 0.998 mm, 0.008%, and 0.999, respectively.

Overview of Memristor-Based Neural Network Design and Applications

Conventional von Newmann-based computers face severe challenges in the processing and storage of the large quantities of data being generated in the current era of “big data.” One of the most promising solutions to this issue is the development of an artificial neural network (ANN) that can process and store data in a manner similar to that of the human brain. To extend the limits of Moore’s law, memristors, whose electrical and optical behaviors closely match the biological response of the human brain, have been implemented for ANNs in place of the traditional complementary metal-oxide-semiconductor (CMOS) components. Based on their different operation modes, we classify the memristor family into electronic, photonic, and optoelectronic memristors, and review their respective physical principles and state-of-the-art technologies. Subsequently, we discuss the design strategies, performance superiorities, and technical drawbacks of various memristors in relation to ANN applications, as well as the updated versions of ANN, such as deep neutral networks (DNNs) and spike neural networks (SNNs). This paper concludes by envisioning the potential approaches for overcoming the physical limitations of memristor-based neural networks and the outlook of memristor applications on emerging neural networks.

The Wako-Saitô-Muñoz-Eaton Model for Predicting Protein Folding and Dynamics.

Despite the recent advances in the prediction of protein structures by deep neutral networks, the elucidation of protein-folding mechanisms remains challenging. A promising theory for describing protein folding is a coarse-grained statistical mechanical model called the Wako-Saitô-Muñoz-Eaton (WSME) model. The model can calculate the free-energy landscapes of proteins based on a three-dimensional structure with low computational complexity, thereby providing a comprehensive understanding of the folding pathways and the structure and stability of the intermediates and transition states involved in the folding reaction. In this review, we summarize previous and recent studies on protein folding and dynamics performed using the WSME model and discuss future challenges and prospects. The WSME model successfully predicted the folding mechanisms of small single-domain proteins and the effects of amino-acid substitutions on protein stability and folding in a manner that was consistent with experimental results. Furthermore, extended versions of the WSME model were applied to predict the folding mechanisms of multi-domain proteins and the conformational changes associated with protein function. Thus, the WSME model may contribute significantly to solving the protein-folding problem and is expected to be useful for predicting protein folding, stability, and dynamics in basic research and in industrial and medical applications.

Connected and automated vehicle control at unsignalized intersection based on deep reinforcement learning in vehicle-to-infrastructure environment

In order to reduce the number of vehicle collisions and average travel time when vehicles pass through an unsignalized intersection with connected and automated vehicle, an improved Double Dueling Deep Q Network method with Convolutional Neutral Network and Long Short-Term Memory is presented in this article. This method designs a multi-step reward and penalty method to alleviate the sparse reward problem using positive and negative reward experience replay buffer. The proposed method is validated in a simulation environment with different traffic flow and market penetration under the mixed traffic conditions of automated vehicles and human-driving vehicles. The results show that compared with traditional signal control methods, the proposed method can effectively improve the convergence and stability of the algorithm, reduce the number of collisions, and reduce the average travel time under different traffic conditions.

Inverse design of coupled subwavelength dielectric resonators with targeted eigenfrequency and Qfactor utilizing deep learning.

Subwavelength all-dielectric resonators supporting Mie resonances are promising building blocks in nanophotonics. The coupling of dielectric resonators facilitates advanced shaping of Mie resonances. However, coupled dielectric resonators with anisotropic geometry can only be designed by time-consuming simulation utilizing parameter scanning, hampering their applications in nanophotonics. Herein, we propose and demonstrate that a combination of two fully connected networks can effectively design coupled dielectric resonators with targeted eigenfrequency and Q factor. Typical examples are given for validating the proposed network, where the normalized deviation rates of eigenfrequency and Q factor are 0.39% and 1.29%, respectively. The proposed neutral network might become a useful tool in designing coupled dielectric resonators and beyond.

Proof of Learning: Towards a Practical Blockchain Consensus Mechanism Using Directed Guiding Gradients (Student Abstract)

Since Bitcoin, blockchain has attracted the attention of researchers. The consensus mechanism at the center of blockchain is often criticized for wasting a large amount of computing power for meaningless hashing. At the same time, state-of-the-art models in deep learning require increasing computing power to be trained. Proof of Learning (PoL) is dedicated to using the originally wasted computing power to train neural networks. Most of the previous PoL consensus mechanisms are based on two methods, recomputation or performance metrics. However, in practical scenarios, these methods both do not satisfy all properties necessary to build a large-scale blockchain, such as certainty, constant verification, therefore are still far away from being practical. In this paper, we observe that the opacity of deep learning models is similar to the pre-image resistance of hash functions and can naturally be used to build PoL. Based on our observation, we propose a method called Directed Guiding Gradient. Using this method, our proposed PoL consensus mechanism has a similar structure to the widely used Proof of Work (PoW), allowing us to build practical blockchain on it and train neutral networks simultaneously. In experiments, we build a blockchain on top of our proposed PoL consensus mechanism and results show that our PoL works well.

Assessing the Relationship between Resource Misallocation and Total Factor Productivity Based on Artificial Neural Network.

For interpretation of China's economy, total factor productivity is considered as one of the crucial aspects which is generally dependent on innovation in technologies especially those driven by both scientific research and efficiency of the methodology or process which is dedicated for the allocation of numerous resources available, among enterprises. It is important to note that various factors, which are either directly or indirectly involved, to cause misallocation of the resources to the enterprises, are overly complex. Therefore, an affective mechanism is needed to be realized which is capable of resolving these issues with the available resources and infrastructures. In this paper, we have focused on the construction or development of an artificial neutral network (ANN) based evaluation model to study the impact of resource misallocation on total factor productivity. Likewise, we have conducted a counterfactual experiment, i.e., simulation only, to thoroughly examine the relationship between two very important factors, that is, (i) resource misallocation and (ii) total factor productivity. To do this, we are aiming at investigating the growth potential of total factor productivity when there is no resource misallocation. After comparing 8 industries in different regions, we conclude that the contribution of capital and labor distortion to total factor productivity is the highest in the eastern region of China with −0.036 and 0.065, respectively, followed by the northeast, central, and western regions. In the experiment, China's total factor productivity growth potential could reach 1.1296, if there is no resource misallocation. The results in this paper would shed some lights on the paths to improve resource allocation efficiency and total factor productivity.

Coconut shell and its biochar as fertilizer amendment applied with organic fertilizer: Efficacy and course of actions on eliminating antibiotic resistance genes in agricultural soil

Biomass amendments have numerous benefits in reducing antibiotic resistance genes (ARGs) in the soil environment. However, there are debatable outcomes regarding the effect of raw biomass and its pyrolytic biochar on ARGs, and the exploration of the influence mechanism is still in infancy. Herein, we investigated the changes in soil ARGs under the organic fertilizer application with coconut shell and its biochar. The results showed that the coconut shell biochar could effectively diminish ARGs, with 61.54% reduction in target ARGs, which was higher than that adding raw coconut shells (p < 0.05). Structural equation modeling indicated that ARGs were significantly affected by changes in environmental factors, mainly by modulating bacterial communities. Neutral community model and network analysis demonstrated that the coconut shell biochar can restrict the species dispersal, thereby mitigating the spread of ARGs. Also, coconut shell biochar exhibited strong adsorption, with a large specific surface area (476.66 m2/g) and pores (pore diameter approximately 1.207 nm, total pore volume: 0.2451 m3/g), which markedly enhanced soil heterogeneity that created a barrier to limit the resistant bacteria proliferation and ARGs propagation. The outcome gives an approach to control the development of ARGs after organic fertilizer application into soil.