Search

At present, low emissivity coating has been widely used in various fields, but damage will greatly reduce the efficiency of low emissivity coating, so the damage detection of low emissivity coating becomes an important work. Based on convolution neural network, a model for automatic identification of coating damage with low emissivity is proposed. Firstly, the optical image data set of low emissivity coating is constructed and extended by means of data enhancement. After that, VGG-19 and ResNet-50 models are built based on tensorflow, and the cross entropy loss function is used in the models. Then, SGD, momentum, RMSprop and Adam are used to optimize the model. In the process of model optimization, the learning rate is adjusted to get the optimal model. The results show that when the learning rate is $5\times 10^{-5}$ and Adam method is used to optimize the model, the recognition accuracy of VGG-19 model is 90.64%, while that of ResNet-50 model is 94.14%. This paper is of great significance for the study of automatic damage identification of low emissivity coatings.

Distributed optimization has become the norm for training machine learning models on large datasets. Learning bigger models on such systems leads to the exchange of large-volume updates. For this reason, limits on network communication in a distributed system can bottleneck learning progress. While compression techniques have been introduced to reduce bit-rates, no methods have yet leveraged the temporal structure that exist in consecutive vector updates. An important example is distributed momentum-SGD where temporal correlation is enhanced by the low-pass-filtering effect of applying momentum. In this paper we design methods that leverage temporal correlation to reduce bit-rates in systems employing momentum-SGD. We demonstrate that a significant reduction in the volume of communication can be realized. Experiments with the ImageNet dataset show that our proposed methods offers up to 40% bit savings compared to widely used methods such as Scaled-sign and Top-K.

Many complicated systems can be represented by complex networks. Their accurate clustering analysis plays a critical role in understanding their intrinsic organizations. An effective Fuzzy-based Clustering Algorithm for Networks (FCAN) has thus been developed. However, its major disadvantage is its slow convergence to optimal or near-optimal solutions. To overcome this problem, we make use of a generalized momentum method to accelerate it and accordingly propose a fast fuzzy clustering algorithm, namely F2 CAN. Experimental results on several practical datasets demonstrate that F2 CAN performed better than FCAN in terms of efficiency while maintaining the same-level accuracy. Hence, it is more promising to conduct an accurate and fast clustering analysis for complex networks.

Nonnegative latent factor (NLF) model has a great ability of acquiring useful knowledge from symmetric, high-dimensional, sparse (SHiDS) matrices. However, the tradition NLF model in terms of the double factorization (DF) technique is somewhat conservative due to the inadequate consideration of difference cases of NLFs. In order to address this issue, we propose an improved DF-SNLF (IDF-SNLF) model to extract NLF matrices in various scenarios. Meanwhile, single NLF-dependent and multiplicative update (SNL-MU) learning method is employed to build an NLF model on SHiDS matrices, yet it suffers from a fairly low convergence rate. So, for the purpose of accelerating the convergence rate, a so-called momentum-based additive gradient descent (MAGD) method is adopted to train the model. Empirical studies on two SHiDS matrices demonstrate that our proposed IDF-SNLF model with MAGD can obtain a desirable performance on both prediction accuracy for missing data and the algorithm convergence rate.

Film cooling is an effective method to achieve reliable thermal protection and reduce skin friction for scramjet combustors. This paper analyzed the effects of incidence angle, blowing ratio and injection distance on flow and heat transfer characteristics of subsonic and supersonic film cooling. The results show that the incidence angle of 30° can achieve optimum cooling effects for both subsonic and supersonic film cooling in the present conditions. And there is a critical blowing ratio (0.8) and momentum ratio (2.0) within the critical distance where the wall-attachment ability of the cooling film suddenly weakens, and the temperature of the inner wall rises, but after this location, the film cooling efficiency increases monotonously with the increase of the blowing ratio and momentum ratio. And the film cooling can significantly increase the surface convective heat transfer coefficient, but decrease the temperature difference between the fluid and the wall more significantly, which results in a decrease of total wall heat flux.

In this paper, we propose and analyze SQuARM-SGD, a communication-efficient algorithm for decentralized training of large-scale machine learning models over a network. In SQuARM-SGD, each node performs a fixed number of local SGD steps using Nesterov's momentum and then sends sparsified and quantized updates to its neighbors regulated by a locally computable triggering criterion. We provide convergence guarantees of our algorithm for general (non-convex) and convex smooth objectives, which, to the best of our knowledge, is the first theoretical analysis for compressed decentralized SGD with momentum updates. We show that the convergence rate of SQuARM-SGD matches that of vanilla SGD. We empirically show that including momentum updates in SQuARM-SGD can lead to better test performance than the current state-of-the-art which does not consider momentum updates., Comment: 58 pages, 8 figures

In this paper, the analysis and design of a 57-64 CMOS Power Amplifier is discussed. The power-combining technique and the capacitor neutralization technique are applied to boost the performance of the purposed PA. The PA is designed in 65nm bulk CMOS process to achieve a saturated output power of 19.4dBm and a peak power-added efficiency of 22%. The power amplifier consumes 300mW from a 1.2V power supply at the output-refereed 1dB compression point of 16.1dBm, and the corresponding power-added efficiency is 13.0%. The passive devices, such as the transformers, the power-combiner and the pads, are designed by Electromagnetic Field Simulation on ADS momentum.

Momentum term based affine projection (AP) algorithm can accelerate the convergence speed of the conditional AP filter at a cost of increased misadjustment. Two schemes are proposed in this paper to improve the performance of the momentum AP algorithm: one is to use an adaptive momentum factor to avoid the confliction between convergence speed and misadjustement; the other combines two momentum AP filters with different momentum factors to achieve the advantages of both filters. The simulation results show that the two proposed algorithms can effectively improve the performance of the momentum AP filter with fast convergence speed and small steady-state error.

In this paper, a modified-momentum stochastic gradient descent (MMSGD) approach for improving the convergence characteristics of active noise control (ANC) system is proposed. MMSGD strategy accelerates the convergence speed of the ANC system while maintaining the low steady-state error. A new update rule and bounds on the learning rate for ensuring the stable working of the proposed MMSGD based ANC system is also derived. A set of simulations on measured plant responses are performed to demonstrates the improved convergence characteristic of the proposed MMSGD based ANC systems over conventional ANC schemes.

many spacecraft using more than 3 reaction wheels for both redundancy and extra maneuverability capabilities. Actuators configuration plays an effective role in the performance of the attitude control system. This paper deals with the management of 4 reaction wheels used simultaneously in the pyramidal configuration. Due to the saturation phenomenon, it is very important to minimizing the angular momentum in the case of using momentum exchange devices. Therefore this paper investigates the best tilt-angle of wheels in pyramidal configuration to perform the maneuver with minimum angular momentum.

Human safety assurance in physical human-robot interactions (pHRIs) has become a challenge for current robotic applications. Considering the safety assurance of a pHRI validation experiment, the viscoelastic properties of a human body part were investigated in a previous study and were added in a pHRI simulation. This study focuses on a threshold momentum value as an injury criterion for momentum control strategy design. A novel safety-oriented control strategy is proposed, which consists of a conventional proportional-derivative (PD) controller, an external momentum observer-based compensator for monitoring the external momentum, and an adjustment loop for shaping the trajectory motion command to bind the external momentum within the injury criterion. The proposed control strategy can be designed without needing to obtain the mechanical properties of the environment. The strategy can also be combined with the related injury criterion. In addition, it shows the effectiveness of ensuring safety even in the worst pHRI clamping situation, where a human body part can escape from such a situation. Simulation results show that the proposed control strategy performs well in the trajectory tracking task, and the adjustment loop successfully shapes the predetermined trajectory motion command to bind the external momentum within the injury criterion.

Stochastic Gradient Descent is a major contributor to the success of the deep neural networks. The gradient provides basic knowledge about the function direction and its rate of change. However, SGD changes the step size equally for all parameters irrespective of their gradient behavior. Recently, several efforts have been made to improve the SGD method, such as AdaGrad, RMSprop, Adam, and diffGrad. The diffGrad is an appropriate and enhanced technique that uses fraction constant based on previous gradient information for gradient calculation. This fraction constant decreases the momentum resulting in slow convergence towards an optimal solution. This paper addresses the slow convergence problem of the diffGrad algorithm and proposed a new adaDiffGrad algorithm. In adaDiffGrad an adoptive batch size is implemented for the diffGrad to overcome the problem of slow convergence. The proposed model is experimented for image categorization and classification over CIFAR10, CIFAR100, and FakeImage dataset. The results are compared with the state of art models, such as Adam, AdaGrad, DiffGrad, RMSprop, and, SGD. The results show that adaDiffGrad outperforms other optimizers and improves the accuracy of the diffGrad.

In this paper, an algorithm of detecting abnormal behavior based on optical flow method for the automatic monitoring is proposed. By comparing the "momentum" of the image and the threshold, the algorithm can determine whether abnormal motion occurs or not. For improving the real-time performance, the algorithm which can be applied to real-time video detection of abnormal behavior, reduces the image resolution, sets the region of interest (ROI) and ignores the tiny changes of optical flow points. Furthermore, calculating accumulating alarm is used instead of single frame image alarm, so the misjudgment is avoided effectively. The feasibility and accuracy of the algorithm are verified through the simulation of abnormal behavior video in elevator.

The theoretical variation of the flywheel angular momentum is given by means of analyzing the attitude dynamics and the disturbance torque of satellite. For the satellite which use magnetorquers for momentum unloading, the change law of the angular momentum has been researched. For formation satellites which require frequent formation control, the change function of the angular momentum under the effect of short-term constant disturbance torque generated by the formation thruster is given, and the need of maximum angular momentum of flywheel under the influence of formation control is obtained. Finally, the maximum time allowed for single-pulse and multi-pulse formation control under the limited angular momentum capability is analyzed, which provides a basis for formation control strategy planning.

As a kind of DC-DC power supply equipment, high-frequency power supply is developing towards digitization, integration and high-power. In the traditional high frequency power supply circuit topology, the inductance parameters at the bridge arm side are easily affected by the change of inductance current, and the DC bus voltage is easily affected by the load fluctuation. In this paper, a high-frequency power supply control method based on inductance current disturbance observer is proposed. The unmeasurable disturbance is estimated from the measurable information in the high-frequency power supply drive circuit, and then the disturbance momentum is compensated in the control algorithm to improve the anti-interference ability and dynamic response speed of the system. Through theoretical analysis and simulation analysis, it is verified that the proposed scheme can effectively improve the anti disturbance ability of high frequency power supply. Considering that the DC-DC power supply is generally nonlinear and intermittent, the total charging current ripple of DC-DC converter circuit working in Buck mode is large, and the DC bus voltage is vulnerable to fluctuations and shocks. This paper presents a DC-DC converter control strategy based on active disturbance rejection observer. The experimental results show that the scheme can reduce the total current ripple, suppress the DC bus voltage fluctuation, and improve the steady-state dynamic characteristics of the control system.

For nonlinear systems with time delays, a generalized predictive control method based on fast regression algorithm (FRA) and bat algorithm (BA) is proposed. Aiming at the model mismatch problem of generalized predictive control, a method for a two-stage neural network model is proposed to establish a nonlinear time-delay predictive model. The fast regression algorithm is introduced to establish the initial neural network model, and then the bat algorithm and the gradient descent with momentum are employed for adjustment of the parameters of the neural network model to obtain the final neural network prediction model. An implicit generalized predictive control (IGPC) is introduced instead of the traditional generalized predictive control (GPC), and this replacement reduces the amount of calculation. The simulation results show that the control performance by our method is better than that via the single neuron PID control method.

Latent factor (LF) model is an effective method for extracting useful knowledge from high-dimensional and sparse (HiDS) data generated by various industrial applications. Parallelized stochastic gradient descent (SGD) is widely used in building a parallelized LF model for handling large-scale HiDS data, but parallelized SGD suffers from slow convergence and considerable time cost. To address this issue, this study incorporates the principle of momentum into parallelized SGD, where momentum decay coefficient and learning rate are adjusted dynamically, and proposes a momentum-incorporated fast parallelized SGD (MFSGD) method to discover latent patterns from large-scale HiDS data. The experiments on two datasets show that the proposed MFSGD method outperforms state-of-the-art parallel SGD methods in terms of computational efficiency.

Today, efforts are made to develop and improve recommendation systems that will direct users to the right product according to their individual preferences during internet shopping. In this study, the recommendation system was designed with Autoencoders, which are one of the methods of deep learning and MovieLens dataset. While designing the system, various optimization algorithms, namely Gradient Descent, Gradient Descent with Momentum, RmsProp and Adam (Adaptive Momentum Optimization), were tried by using TensorFlow in the Python programming language. Moreover, the effect of increasing the amount of the data on the optimization algorithm was analyzed. Consequently, it was effectively demonstrated that the most successful one was the Adam algorithm with a test error of 1.363. It was also observed that decreasing the sparsity on the training data leads to a lower test error.

Two numerical wake model were established, one based on full geometry turbine and one based on actuator disc theory. The Actuator disc momentum theory assumes turbine as a constant sink of momentum in numerical domain. Results of numerical models were compared with experimental results, the experimental data is in good agreement with the simulation data. The difference between two models were analyzed in detail. The the difference between simulation results of actuator disc and test results are slightly larger in near wake region, the difference gradually decreased as the axial distance increased. The lack of the simulation of rotation of the model may lead this difference. The effect to wake recovery of blockage ratio was studied using numerical model based on actuator disc theory, 11 cases were studied, the results showed that the blockage ratio can heavily effect the recovery speed of turbine wake. Along the centerline, the degree of wake recovery decreased as the blockage ratio increased before X/D=10.8, and the degree of wake recovery increased as the blockage ratio increased after X/D=10.8.

Mesoscale eddies play an essential role in ocean momentum, energy, heat, and mass transport. Thus, it can also significantly affect the distribution of nutrients, phytoplankton, as well as the shoal of fish. In order to quickly and accurately locate the mesoscale eddies in the ocean, we fused the velocity, divergence, vorticity, and Okubo-Weiss (OW) parameter information with the sea surface height anomalies (SSHA) data, respectively. Then, the feature pyramid network (FPN) in the deep learning method was adopted to extract the eddy position from fused information. The experiment shows that the proposed method can effectively detect the mesoscale eddy locations. Besides, it can also be concluded that, compared with the merged data with velocity and divergence, the FPN performs better on fused data with vorticity and OW information.

The snow cover on sea ice is an important parameter controlling heat and momentum fluxes in our polar regions. Our understanding of snow thickness distributions on sea ice is severely limited by its vastness and numerous logistical difficulties. As such, we rarely collect enough in situ data from similar geographic locations to determine if and how the snow thickness distribution changes spatiotemporally. Geophysical changes in snow cover manifest as differences in dielectric properties, which are detectable in microwave emission and backscatter. Active microwave remote sensing offers improved spatial resolution when compared to passive microwave approaches. We apply our recently developed method that exploits the indirect thermodynamic control of the snow cover on near ice surface geophysical properties. The variance of C-band (5.3 GHz HH-polarization) microwave backscatter in winter (prior to melt) is assessed and is then used to estimate relative snow cover thickness and distribution. We assess the capability of our approach over landfast, first-year sea ice in the Canadian Arctic Archipelago and evaluate and compare our method against the Canadian Regional Ice Ocean Prediction system and AMSR2 passive microwave snow thickness estimates. Results demonstrate that this method can separate thick snow from thin snow on thick FYI within a thickness range of 5 to 45 cm at a spatial resolution of less than 5 km.

This work demonstrates the first implementation of lithium niobate-on-silicon (LNoSi) non-reciprocal acoustic delay lines (ADL) with tunable insertion loss (IL) utilizing the acoustoelectric (AE) effect. Due to the AE effect, the direction-and the intensity-dependent momentum exchange between the drifting electrons in the Si layer and the acoustic phonons can be utilized to break the intrinsic reciprocity of the ADLs in order to control their frequency response. A 5.2 dB improvement in the IL and a 14.2 dB increase in the reverse isolation (i.e. a 19.4 dB non-reciprocal transmission ratio) is achieved through injecting a 400 µA current in one of the ADLs presented here. This opens up possibilities of merging long delays, tunable attenuators, and switches in a single miniaturized device which is a critical stepping stone in fulfillment of full-duplexed microwave systems.

In order to make tide prediction more efficient and accurate, this paper proposes a method based on adding momentum term to optimize wavelet neural network for modular tide prediction. In the study, the tides are divided into astronomical tides which are affected by the tidal force of celestial bodies and non astronomical tides which are affected by environmental factors and other factors. The harmonic analysis method and wavelet neural network are used to predict the tides respectively. When using wavelet neural network to predict non astronomical tide, the method of adding momentum term is used to optimize the weight and parameters of wavelet neural network to improve the prediction accuracy. The simulation experiment is carried out by using the data measured at Tampa tide station in western Florida. The results show that the tide prediction model can improve the prediction accuracy and is efficient and feasible.

This paper proposes a momentum compensation adaptive filter based on variable parameters. Aiming at the traditional adaptive filtering method, the problems of slow convergence and poor real-time performance are considered. The momentum term was added to the expression of the adaptive filter parameter update to compensate, and the algorithm step size was changed to adaptive form. Finally, the active vibration isolation verification was performed on the Stewart platform. The results show that this method can obtain better convergence speed. Especially for active vibration isolation systems with conventional vibration sources, this method has better vibration isolation effects.

We propose a method that adapts hyperparameters, namely step-sizes and momentum decay factors, in on-line learning with classic momentum and ADAM. The approach is based on the estimation of the short- and long-term influence of these hyperparameters on the loss value. In the experimental study, our approach is applied to on-line learning in small neural networks and deep autoencoders. Automatically tuned coefficients surpass or roughly match the best ones selected manually in terms of learning speed. As a result, on-line learning can be a fully automatic process, producing results from the first run, without preliminary experiments aimed at manual hyperparameter tuning.

Wind speed is a key factor for wind energy generation and air pollution dispersion. However, current accuracy of short-term wind speed prediction is poor and leads to economic loss in wind power system. Therefore, improving the accuracy is urgent. A stack long short-term memory (LSTM) network with fully connected network was adopted in this study to predict wind speed. Rather than just using the related meteorological variables (e.g. wind direction, temperature, relative humidity, dew point temperature, pressure, etc.) and wind speed sequences, we also added the convection terms of the momentum equations. The mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE) and mean absolute percentage error (MAPE) were used to evaluate the prediction’s performance. The results showed our new proposed method could obtain better accuracy than previous studies.

Stochastic gradient descent(SGD) is the fundamental sequential method in training large scale machine learning models. To accelerate the training process, researchers proposed to use the asynchronous stochastic gradient descent (A-SGD) method in model learning. However, due to the stale information when updating parameters, A-SGD converges more slowly than SGD in the same iteration number. Moreover, A-SGD often converges to a high loss value and results in lower model accuracy. In this paper, we propose a novel algorithm called Trend-Smooth which can be adapted to the asynchronous parallel environment to overcome the above problems. Specifically, Trend-Smooth makes use of the parameter trend during the training process to shrink the learning rate of some dimensions where the gradients' directions are opposite to the trends of parameters. Experiments on MNIST and CIFAR-10 datasets confirm that Trend-Smooth can accelerate the convergence speed in asynchronous training process. The test accuracy that Trend-Smooth achieves is shown to be higher than other asynchronous parallel baseline methods, and is very close to the SGD method. Moreover, Trend-Smooth can also be combined with other adaptive learning rate methods(like Momentum, RMSProp and Adam) in the asynchronous parallel environment to promote their performance.

Direct aperture optimization (DAO) is an effective method to generate high-quality intensity-modulated radiation therapy treatment plans. In generic DAO, the direction of negative gradient descent is generally used to determine the aperture shape. However, this strategy can reduce the convergence rate, especially near the optimal value. We propose aperture shape generation based on the direction of gradient descent with momentum, where column generation is implemented as carrier. During aperture shape generation of column generation, the current aperture gradient map is first calculated. Then, the gradient with momentum is calculated based on the existing gradient information. Finally, the direction of gradient descent with momentum is constructed for obtaining the deliverable aperture shape by solving the pricing problem. To verify the effectiveness of the proposed method, we conducted comparative experiments on two head and neck and two prostate tumor cases. Compared with generic column generation, the proposed method can effectively protect the organs at risk while ensuring the required dose distribution to the target. Using the proposed method, the number of apertures and optimization time can be reduced by up to 30.95 and 32.96%, respectively, compared to the conventional approach. The experimental results suggest that the proposed method can accelerate the search speed and improve the quality of treatment plans.

The demand for recommender systems in E-commerce industry has increased tremendously. Efficient recommender systems are being proposed by different E-business companies with the intention to give users accurate and most relevant recommendation of products from huge amount of information. To improve the performance of recommender systems, various stochastic variants of gradient descent based algorithms have been reported. The scalability requirement of recommender systems needs algorithms with fast convergence to generate recommendations of specific items. Using the concepts of fractional calculus, an efficient variant of the stochastic gradient descent (SGD) was developed for fast convergence. Such fractional SGD (F-SGD) is further accelerated by adding a momentum term, thus termed as momentum fractional stochastic gradient descent (mF-SGD). The proposed mF-SGD method is shown to offer improved estimation accuracy and convergence rate, as compared to F-SGD and standard momentum SGD for different proportions of previous gradients, fractional orders, learning rates and number of features.

A multistage attitude control and momentum management (ACMM) control strategy is proposed for space stations. The ACMM, which includes closed-loop angular momentum feedback (CAMF) and torque equilibrium attitude (TEA) tracking, is achieved using the gravity gradient and an aerodynamic torques. The CAMF is employed as the normal flight mode, and TEA tracking is used to unload the angular momentum after attitude stabilization. Hence, the angular momentum imparted by the attitude control device does not need to be unloaded using thrusters during the entire operating phase. The system performances, especially the relationship between the environmental torques and the attitude, are analysed. The internal model principle is utilized to suppress the attitude fluctuation during the CAMF mode, and a quintic polynomial is utilized to realize TEA tracking. The pole placement algorithm, which can obtain the feedback matrix in the linear quadratic regulator (LQR) without choosing a weight matrix, is introduced to design the LQR controllers. This approach is demonstrated for a future Chinese space station, and the simulation results verified the effectiveness of the proposed control algorithm.

Anintrusion detection system is one of the methods that help when intrusion is detected to determine scheme safety. In this paper, NSL-KDD data set efficiency is assessed using enhanced RNN with momentum gradient descent and adaptive learning rate (GDX). Before destabilizing the core network, an intelligent IDS should be used to achieve this goal, and we propose a new RNN-GDX algorithm.Based on multiple performance measures, results are analyzed and better accuracy has been discovered. We achieved minimum (MSE) mean square error rate and maximum accuracy Using the RNN-GDX algorithm for NSL-KDD data set results showed that RNN-GDX learned better as well as overall efficiency isincreased to 93.3 percent & 97.68 percent respectively, with 29 input and 29 hidden layer neurons and 41 input & 41 hidden layer neurons changing to the lowest value of 0.01.

As an important parameter that characterizes the structure of Atmospheric boundary layer (ABL), the boundary layer height (BLH) determines the degree of turbulent diffusion and exchange of heat and momentum within ABL. In this paper, synchronized observations of ground-based microwave radiometer (MWR) and radiosonde in Beijing Southern Meteorological Observatory from April 2017 to March 2018 are obtained to estimate BLH, based on the multiple linear regression algorithm and bulk Richardson number method, respectively. The BLHs derived from the two observations are compared and the temporal variation of BLH in Beijing are finally analyzed. Main conclusions are as follows: 1) the BLH retrieved by MWR is higher than that obtained by radiosonde by 157 m on average. The average deviation of BLH between MWR and radiosonde is smaller at 08 Beijing time (BJT) and 20 BJT than at 14 BJT, and the deviation is relatively smaller in spring and summer than other seasons. 2) BLH at 08 BJT is generally lower than that at 14 BJT and 20 BJT from both MWR and radiosonde observations. 3) radiosonde-obtained BLH peaks in the May while MWR-retrieved BLH reaches maximum in January and February. 4) BLH in Beijing at night is relatively low, which reaches the minimum $(\sim 622\mathrm {m}$ on average) around 05 BJT or 06 BJT. BLH gradually increases after 08 BJT and reaches the maximum $( \sim 908\mathrm {m}$ on average) around 15 BJT.

The ARM-based server processors have been gaining momentum in high performance computing (HPC). While not designed specifically for HPC, Huawei Kunpeng 916 processor has 32 ARMv8 cores and is tempting for HPC workloads. However, its potential remains unknown. To throughly understand the potential, we conducted a systematic evaluation in three steps by using: 1) three well-known benchmarks (HPL, STREAM, and LMbench); 2) three typical scientific kernels (SpMV, N-body, and GEMM); 3) three widely used mini-apps (TeaLeaf, Neutral, and SNAP) and a real-world application GTC-P. We compared the performance results of Kunpeng 916 with that of Intel Xeon E5-2680v3/4 (Haswell/Broadwell). The evaluation results show that Kunpeng 916 has higher memory bandwidth than the two Intel processors, thus it can achieve compelling performance for running memory bound HPC applications.

Rapid urban growth has made the traffic predictions difficult which creates worst situations on the road. These days all the major roads of the various cities are becoming congested all over the world. Around 43% of the people in Hyderabad complain about the congestion every day. The congestion can be reduced with the advent of precise forecast of the traffic flow, therefore the need of short-term traffic prediction gains momentum these days. The study is based on short term traffic prediction using Seasonal Autoregressive Integrated Moving Average (SARIMA) and the Facebook PROPHET package in R software. A PROPHET model and SARIMA model was developed to analyze short term traffic prediction using the seven day hourly volume count of the National Highway 744. The best SARIMA model ARIMA(1,0,0)(2,0,0) [24] was developed to forecasts traffic volume and the accuracy of the model is checked in terms of mean absolute percentage error(MAPE) and root mean square error(RMSE). From the PROPHET model we will be the getting the forecast trend with the boundary limits for the forecasted date.

In the case of existing information diffusion models, probability of information diffusion only follows parameters of neighbor nodes sending and receiving information. But the momentum of diffusion gains in an exponential fashion whenever the information passes through influential nodes in a real network. For example, when information diffuses through SNS, the speed of diffusion increases whenever information passes nodes that is influential users. This paper quantifies the degree of influence as information vitality, and proposes the model using it. In addition, the paper considers relationship between diffusion speed and information vitality in simulation of the model.

This paper aims to propose a nonlinear attitude controller for a drone system controlled by the control momentum gyros (CMGs). The dynamic modeling and the control mechanism of the considering system are first investigated. After that, based on the assumption of the timescale separation, the proposed controller is designed using the feedback linearization methodology. To make the resultant controller follow the three-loop structure, specific forms of the desired error dynamics are utilized in the design step. The characteristics of the proposed method are investigated through numerical simulations. The results show that the proposed controller can provide a good tracking performance against model uncertainties as well as model nonlinearities.

The security of information system oriented to network users is very important. The deployment scheme of information system based on cloud service has a rapid development momentum with appropriate cost and high security. Take the security protection deployment of an Alibaba cloud platform as an example to introduce the concrete deployment and implementation of the security protection of information system deployed on cloud servers. After the security protection is deployed and running, the information system runs normally. The number of detected malicious attacks has been significantly reduced. The system security is proven to be guaranteed.

In order to study the effect of wall insulation conditions and different airflows on burning rate of pool fires, a small-scale ethanol pool fire was conducted experimentally under three pool diameters and a certain lip height in a wind tunnel. The results showed that the non-monotonic mass loss rate decreases first and then increases with the increase of crosswinds at a certain lip height. The burning rate increases with the increased pool diameter of the non-insulated pool fires, while an opposite trend of burning rate was found in insulated pool fires at the same sizes. The dimensionless characteristics of Richard number presents the variation relationship between the buoyancy force and momentum during different airflows. The insulated liquid pool showed a higher burning rate than that in a non-insulated pool with the characteristic parameter of reciprocal of Richard number less than 2. With the value of Richard number continue increased, the burning rate in the insulated pool fires decreased due to the difference between the increased crosswind and heat transfer. The mass loss rate of pool fires without insulation increased with the increase of pool diameter, while an opposite trend was found in insulated pool fires at the same size. The experimental data supported to establish a ratio coefficient of burning rate with or without insulation under different crosswinds.

This paper presents a new technique namely Grey Wolf Optimizer- Artificial Neural Network (GWO-ANN) as a technique to forecast electrical load. GWO is a meta heuristic technique inspired by the hierarchy of leadership of the grey wolf hunting mechanism in nature. Four types of grey wolves such as alpha, beta, delta, and omega are employed for simulating the leadership hierarchy. In addition, the three main steps of hunting, searching for prey, encircling are also imitated in the algorithm. GWO is utilized to determine the optimal momentum rate and learning rate of ANN for accurate prediction. In the ANN configuration, the temperature, humidity, wind speed, maximum power, and average power were used as the input data. While total power was used as the output data. ANN is trained by adjusting the parameters of momentum rate and learning rate until the output data matches the actual data. The performance of GWO-ANN was compared to the performance of ANN and Particle Swarm Optimization - Artificial Neural Network (PSO-ANN). The results showed GWO-ANN provide better result in terms of the Mean Absolute Percentage Error (MAPE) and coefficients of determination (R2) as compared to other methods.

Stochastic Gradient Descent (SGD) is a well-received algorithm for large-scale optimization in neural networks for its low iteration cost. However, due to Gradient variance, it often has difficulty in finding optimal learning rate and thus suffers from slow convergence. Using momentum is proven to be a simple effective way of overcoming the slow convergence problem of SDG as long as momentum is properly set. According to the performance metrics, this paper proposes a novel statistical model for analyzing the performance of neural networks. The model takes into account learning rate and momentum, and the method can be used to evaluate and verify their interaction effects on neural network performance. Our study shows that the interaction effects are significant. When momentum has a value smaller than 0.5, the impact on the training time is not statistically noticeable.

High-dimensional and sparse (HiDS) matrices are commonly encountered in many big-data-related industrial applications like recommender systems. Latent factor (LF) analysis via stochastic gradient descent (SGD) is greatly efficient in discovering latent patterns from them. However, as a sequential algorithm, SGD suffers considerable time cost and low scalability when handling large-scale problems. To address these issues, this study proposes parallelized, momentum-incorporated stochastic gradient descent (PMSGD) scheme, which incorporates momentum effects into an SGD scheme as well as implementing its parallelization via careful data splitting. Based on a PMSGD method, we achieve a PMSGD-based LF (PLF) model to execute fast LF analysis on HiDS matrices from a recommender system. Experimental results on two HiDS matrices arising from industrial applications indicate that owing to the careful design of PMSGD, a PLF model outperforms state-of-the-art parallel LF models significantly in terms of computational efficiency.

Quality-of-service (QoS) of Web services vary over time, making it a significant issue to discover temporal patterns from them for addressing various subsequent analyzing tasks like missing QoS prediction. A Latent factorization of tensors (LFT)-based approach proves to be highly efficient in addressing this issue, which can be built through a stochastic gradient descent (SGD) solver efficiently. However, an SGD-based LFT model frequently suffers low-tail convergence. For addressing this issue, we present a momentum-incorporated latent factorization of tensors (MLFT) model, which integrates a momentum method into an SGD-based LFT model, thereby improving its convergence rate as well as maintaining the prediction accuracy for missing QoS data. Empirical studies on two dynamic industrial QoS datasets show that compared with an SGD-based LFT model, an MLFT model achieves faster convergence rate and higher prediction accuracy.

With the development of intelligent manufacturing, image processing technology is becoming more and more close to life. Image processing measurement gains momentum development in the field of ITS. Vehicle detection system aims at obtaining data of traffic, length of models of vehicle. Virtual detection loop system is poor in accuracy,the author of this paper puts forward the improvement of Kalman algorithm, improving the accuracy of vehicle detection.At the same time,reducing false detection rate of vehicle information.

Development of solar power all around the world has gained momentum recently. Like other renewable energy recourses such as hydro and wind, fortunately, Tajikistan is equally endowed with the solar energy resources. The study is based on “Bokhtar, Qurghonteppa” a region located in Tajikistan. The experiment was conducted to explore the solar potential of the region by using daily average solar irradiance data recorded for a period of 1 year. For this study the performance of Dual Axis Solar Tracker (DAST) and Static Solar System (SSS) was assessed and compared. Apart from measuring daily and monthly average solar intensity records by using DAST and SSS, additionally, seasonal and yearly energy gain and efficiencies were also calculated. Results obtained suggest that, having an overall efficiency advantage of 44.02 % during one year period, DAST can achieve an efficiency as high as 80.57% during autumn season only.

More and more GEO Satellites adopts electric thrusters to execute station-keeping and momentum dumping. Profiting from the high specific impulse of electric propulsion, there is substantially decrease in fuel consumption and improvement in payload capacity. Quantity and configuration of electric thrusters have a large influence in station-keeping efficiency and momentum dumping capacity. Considering the necessary of system backup, the station-keeping and momentum dumping must be performed correctly while at least one thruster breaks down. In this paper, a method of station-keeping and momentum dumping capacity analysis is introduced, from which the minimum configuration of electric thrusters is induced. In addition, the minimum configuration of electric thrusters for GEO satellite station-keeping and momentum dumping with backup is discussed. At last, the relationship between thrusters' installation parameters and capacity of station-keeping and momentum dumping is analyzed. In view of the above, the designing method of electric thrusters' installation parameters is obtained.

Spin-transfer torque random access memory (STT-RAM) is gaining momentum as a promising technology for high density and embedded nonvolatile memory. To enable the design of STT-RAM circuits for memory and computing, there is a need for accurate compact models capable of predicting the stochastic behavior. Here, we present a detailed model accounting for the anomalous thermal regime of switching deviating from the thermal model below 200 ns. Anomalous switching is explained by the non-linear barrier lowering of the perpendicular magnetic anisotropy (PMA). The model is validated against write error rate (WER) data and applied to the design of random number generator (RNG) and stochastic computing primitives with STT-RAM.

The momentum-based stacked attention networks (SANs) is one of the best models for image question answering. However, we find that it is easy to fall into the local optimal solution, which results in the higher question answering error rate. To solve the problem, we propose adaptive bold driver (ABD). The experimental results and analysis show that it outperforms the state-of-the-art global learning rate adaptive algorithm in the local learning rate adaptive stochastic gradient descent (SGD). It is deeply integrated with momentum, and we propose momentum based on ABD (MABD). The experimental results show that its accuracy is 2.33% higher than the baseline (momentum), 2.54% higher than momentum based on bold driver, and 1.80% higher than the annealing-based momentum. The experimental analysis proves that it is the state-of-the-art optimization algorithm in the SANs-based image question answering and it has effectiveness, significance, generalization performance, and promotional value.

This paper applies an algorithm to detect the residual momentum of a two-link flexible manipulator to estimate the fault state. First, we establish a dynamic model based on the Lagrange equations and assumed modes method. For the two-link flexible manipulator in the vertical plane, we consider both the dead weight of arms and the weight of loads. Second, we get residual momentum based on the derivative of momentum. Using the dynamic model of the above manipulator, we deduce the residual momentum fault detection algorithm. Finally, MatlablSimulink is used to establish simulation platform, and the influence of flexible deformation mode on the rotational angle in the calculation of residual momentum is taken into account. The simulation is about both drives faults and collision faults, then gives out the change curves of the corresponding residual momentum. The results show that the residual momentum model established in this paper can be applied to rigid arms and flexible arms. According to the size and trend of residual momentum, we can carry out corresponding fault classification and diagnosis. And from the simulation, it shows that the residual momentum reflects the value of the desired input torque in that period of time.

We introduce the concept of Momentum Driven Structures (MDS) made of inertially actuated units linked together by compliant elements as a potential solution for rough environments exploration. We propose a control method for MDS based on the bio-inspired concept of Central Pattern Generator (CPG) and study in simulation the impact of compliance distribution on locomotion performance using population based optimization techniques. Our results suggest that compliant structures outperform their rigid counterparts in terms of distance traveled. In addition, we show that co-evolved structures perform only marginally better than their control-only optimized equivalent, highlighting the fact that compliance modulation may not be a significant asset in such experiments, considering the related hardware complexity it introduces.

Meteor Radar is widely used to study Gravity wave (GW) variances and their Momentum Fluxes (MFs) at mesospheric altitudes. However, the accuracy of meteor radar observations of GWMF should be evaluated before they can be used for research applications. In this regard, the present study evaluates the meteor radar observations of GWMF obtained from Thumba [8.5°N, 77°E (2006- 2015)], Kototabang [0.2°S, 100.3°E (2002-2017)] and Tirupati [13.63°N, 79.4°E (2013-2018)]. These radars provide hourly zonal and meridional wind observations round the clock in 80-100 km height domain known as mesosphere lower thermosphere region. Three methods are employed to evaluate the meteor radar performance in estimating the simulated GWMFs, Viz., 1) Day by day analysis, which is traditional Hocking method 2) Composite day analysis and 3) Modified Composite day analysis. After evaluating these methods by employing GW simulations, it is found that modified composite day analysis shows good agreement and the same is used for constructing the GWMF climatologies over the observational locations. The significance of the present study lies in evaluating the three different methods for estimating GWMF by employing GW simulations and meteor radar observations.