Large Amount Of Parameters Research Articles

Improved lightweight federated learning network for fault feature extraction of reciprocating machinery

Abstract The working environment of reciprocating machinery is complex, characterized by nonlinear and non-stationary signals. Deep learning can solve the above problems, but it has its own problems such as complex model and large amount of parameters. Additionally, privacy considerations among enterprises prevent data sharing, leading to the emergence of ‘data islands’ and inadequate training of deep learning models. Based on the above analysis, this paper proposes a reciprocating mechanical feature extraction method based on an improved federated lightweight network. A lightweight network SqueezeNet model is used to solve the problems such as long training time of deep learning. By establishing a federated learning framework, the reciprocating mechanical data can be collectively diagnosed across various enterprises, thereby addressing the problem of limited model training caused by insufficient data. Furthermore, to enhance the accuracy of network training and diagnosis, modifications are made to the SqueezeNet network to reduce the number of model parameters while increasing the number and variety of feature extractions. Experimental results demonstrate that when the number of 1 × 1 and 3 × 3 channels is 1–7, the fault diagnosis accuracy is the highest, about 97.96%, which enriches the categories of feature extraction. The number of parameters in In-SqueezeNet is 56% of that in SqueezeNet network model, and the training time is reduced by nearly 15%. The fault diagnosis accuracy is increased from 95.1% to 97.3%, and the diversity of extracted features is increased. Compared with other network models such as ResNet, the improved lightweight federated learning network has a fault diagnosis accuracy of 96.6%, an improvement of 10.6%. At the same time, the training time was reduced to 1982s, a reduction of about 41.5%. The validity of the proposed model is further verified.

A Tiny Model for Fast and Precise Ship Detection via Feature Channel Pruning

It is of great significance to accurately detect ships on the ocean. To obtain higher detection performance, many researchers use deep learning to identify ships from images instead of traditional detection methods. Nevertheless, the marine environment is relatively complex, making it quite difficult to determine features of ship targets. In addition, many detection models contain a large amount of parameters, which is not suitable to deploy in devices with limited computing resources. The two problems restrict the application of ship detection. In this paper, firstly, an SAR ship detection dataset is built based on several databases, solving the problem of a small number of ship samples. Then, we integrate the SPP, ASFF, and DIOU-NMS module into original YOLOv3 to improve the ship detection performance. SPP and ASFF help enrich semantic information of ship targets. DIOU-NMS can lower the false alarm. The improved YOLOv3 has 93.37% mAP, 4.11% higher than YOLOv3 on the self-built dataset. Then, we use the MCP method to compress the improved YOLOv3. Under the pruning ratio of 80%, the obtained compressed model has only 6.7 M parameters. Experiments show that MCP outperforms NS and ThiNet. With the size of 26.8 MB, the compact model can run at 15 FPS on an NVIDIA TX2 embedded development board, 4.3 times faster than the baseline model. Our work will contribute to the development and application of ship detection on the sea.

Research on Intelligent Fault Diagnosis of Rolling Bearing Based on Improved ShufflenetV2-LSTM

The timing characteristics of the fault vibration signal of the rolling bearing are ignored by the ShufflenetV2 network. For the bearing fault problem of multiple working conditions, the fault diagnosis signal is extracted by the feature, which cannot be performed efficiently and accurately. The ShufflenetV2 network has a deep number of layers and a large amount of parameters, which causes the network to be prone to overfitting. In response to the above problems, an improved ShufflenetV2-LSTM intelligent fault diagnosis system is proposed, which is a model in which the long short-term memory network (LSTM) layer and the Dropout layer are serially embedded in the ShufflenetV2 network. This method preserves the ability of the ShufflenetV2 network to extract features, and the advantage of the ability of LSTM to strengthen the data sequence is also inherited, so this improves the accuracy of fault diagnosis. The generalization ability of the model can be enhanced by Dropout, which can effectively suppress the degree of overfitting. In addition, this paper also optimized the activation function and optimizer in the model to make up for the additional time cost brought by the Dropout layer, so that the robustness of the system is improved and fault diagnosis can be analyzed efficiently. Experimental analysis shows that the diagnosis accuracy of the proposed algorithm is as high as 97% and early failures of rolling bearings can be effectively identified in real time.

Constitutive rubber model suitable for rolling resistance simulations of truck tyres

Tyres are a vital vehicle component forming an interface between a vehicle and the road, enabling the generation of braking, steering and traction forces. However, they also generate rolling resistance which researchers have tried to minimise through the years for environmental and economic reasons. Despite numerous attempts to model rolling resistance of tyres there still does not seem to exist a simple, flexible and accepted way of modelling rolling resistance in the time domain as well as parametrising models in an easy and accessible way. This study explores a simple and intuitive way of parametrising a hyperviscoplastic parallel rheological framework. In the experimental part of this study, rubber samples with various amounts of carbon black filler are extracted from a truck tyre section and tested using dynamic mechanical analysis. The test data was used to parametrise the material model. The model consists of Mooney-Rivlin hyperelasticity, 40 Prony elements and 8 perfectly plastic elements with Ogden hyperelasticity. The paper introduces a method to obtain a large number of parameters using only six tuneable parameters, which simplifies the tuning of the model drastically. The parametrised model is suitable for tyre rolling resistance simulations with frequency and strain amplitude dependency of the storage and loss modulus. A wide range of strain amplitudes and frequencies can be covered with the proposed method and it is possible to achieve a good fit for the storage and loss modulus values with the benefit of only a few tuneable parameters. Additional Prony or plastic networks do not increase the amount of tuneable parameters. Moreover, the method can be used to parametrise the material using manual iterations which is generally not possible for a parallel rheological framework with such a large amount of parameters.

Vehicle detection algorithm based on LW-SSD

Due to the complex structure of the traditional SSD network model and the large amount of parameters, it is not suitable for porting to mobile or embedded platforms, so the LW-SSD vehicle detection algorithm is proposed, this paper redesigns the basic feature extraction network of SSD, using MobileNet feature extraction network instead of VGG16, for the original MobileNet to improve, remove the original MobileNet network average pooling layer and Softmax layer, while reducing the MobileNet network from the original 28 layers to 13 layers. Then, on the basis of the subsequent network, feature maps of different scales are scaled, residual mapping and deep feature fusion operations, and the low-level position information and high-level semantic information are effectively combined to obtain new feature maps of different scales. The designed network model is trained using the Tensorflow framework, and the test effect of the network model is evaluated. The experimental results show that the accuracy of the network model designed in this paper is 78.76%, which is 9.27% lower than the original SSD network model. However, the amount of parameters is only 1/4 of the SSD network model. At the expense of partial accuracy, the design of a lightweight network model is exchanged, which lays the foundation for subsequent development on mobile or embedded platforms.

Research and Implementation of Breast Cancer Intelligent Recognition Algorithm Based on Deep Convolutional Neural Network

Breast cancer is one of the most dangerous cancers with a particularly high mortality rate for women. In the face of thousands of female pathological pictures, how to use computer related control algorithms to quickly and efficiently intelligently idently female breast cancer images has become an urgent problem in the world. With the development of computer vision and deep convolutional neural networks, this paper applies the convolutional neural network (CNN) algorithm to solve the classification and recognition computer tasks of histological images for breast cancer, aiming to learn from tens of thousands of histological images features for breast cancer, and to assist doctors in diagnosing patients’ disease was found by computer control algorithm. There are many well-known neural network computer algorithms, such as AlexNet, Inception-Net, ResNet. Image recognition for breast is a binary classification problem, and it is not appropriate to use a ResNet network with a large amount of parameters. Therefore, this paper borrowed Alexnet algorithm and redesigned the algorithm, then used it for breast cancer experiments. We compared with the capsule algorithm used by Anupama authors. Our experiment improves the accuracy rate by 2%, and reduces the amount of neural network parameters, which greatly speeds up the training time.

Dual attention convolutional network for action recognition

Action recognition has been an active research area for many years. Extracting discriminative spatial and temporal features of different actions plays a key role in accomplishing this task. Current popular methods of action recognition are mainly based on two-stream Convolutional Networks (ConvNets) or 3D ConvNets. However, the computational cost of two-stream ConvNets is high for the requirement of optical flow while 3D ConvNets takes too much memory because they have a large amount of parameters. To alleviate such problems, the authors propose a Dual Attention ConvNet (DANet) based on dual attention mechanism which consists of spatial attention and temporal attention. The former concentrates on main motion objects in a video frame by using ConvNet structure and the latter captures related information of multiple video frames by adopting self-attention. Their network is entirely based on 2D ConvNet and takes in only RGB frames. Experimental results on UCF-101 and HMDB-51 benchmarks demonstrate that DANet gets comparable results among leading methods, which proves the effectiveness of the dual attention mechanism.

Local image quality measurement for multi-scale forensic palmprints

Numerous studies show that palmprint image quality has a significant effect on every stage of a palmprint recognition system. Although some palmprint image quality measurement(PIQM) methods are proposed, some insufficiency in classification accuracy occurs and attention to detail in measuring local area image quality of multi-scale palmprint images is lacking. On the one hand, the classification accuracy is not very high for 2-class classification and it degrades significantly as the number of classes increases. On the other hand, local area image quality measurement of multi-scale palmprint images has not yet been resolved since the handcrafted features designed through domain knowledge usually works for certain scale image blocks. Meanwhile, the intricate domain knowledge used in the previous methods is difficult for some common users to acquire. In this paper, we propose an end-to-end deep-learning method of strengthening representation ability that learns more abstract, essential, and reliable features to measure the local image quality for multi-scale forensic palmprints. Popular convolutional neural networks (CNNs) are considered because of their powerful representation ability in learning complex features. However, the powerful existing CNNs usually have complex architectures with a large amount of parameters, which need the support of high-performance computers. They are not suitable to be used directly for palmprint image quality assignment and the follow-up palmprint recognition work, which prefers real-time response on commonly available personal computers or even mobile devices. Hence, a new lightweight CNN must be designed to achieve a trade-off between high classification accuracy and practical usability. Considering the attributes of under-processed input images, we reduce the weight of the CNN architecture by reducing the amount of some parameters, and finally a lightweight CNN is designed. As a result, a raw rectangular palmprint image of variable size can be put into the trained model directly and a quality label quickly predicted with high accuracy. After comparison with previous methods, results show that the proposed method can deal with un-pre-processed raw images of a multi-scale input size. Furthermore, it can acquire a richer amount of quality classes with a higher accuracy, which are stable on many different datasets. It also leads to finer and more precise full palmprint image quality maps when compared to previous methods.

Distributed Gauss-Newton Optimization with Smooth Local Parameterization for Large-Scale History-Matching Problems

Summary Finding multiple posterior realizations through a reservoir history-matching procedure is of significance for uncertainty quantification, risk analysis, and decision making in the course of reservoir closed-loop management. Previously, an efficient, distributed, optimization algorithm—global linear-regression (GLR) with distributed Gauss-Newton (GLR-DGN)—has been proposed in the literature to iteratively minimize multiple objective functions by performing Gauss-Newton (GN) optimizations concurrently while dynamically sharing information between dispersed regions in a reduced parameter space. However, theoretically, the required number of initial training models should be larger than the number of parameters to guarantee a unique solution for the GLR equation for sensitivity-matrix estimation. This limitation makes the large-scale reservoir history-matching problem with a large amount of parameters almost intractable. We enrich the previous history-matching framework by integrating our recently proposed smooth local parameterization (SLP) and DGN for the sensitivity-matrix calculation. Motivated by the fact that one specific flow response mainly depends on a few influential or local parameters, which can be generally identified by the physical position of the wells (e.g., the parameters in the zone surrounding the wells), which is particularly true for large-scale reservoir models, this paper presents a new integration of subdomain linear-regression (SLR) with DGN, referred to as SLR-DGN. This SLP allows us to independently represent the globally spatial parameter field within low-order parameter subspaces in each subdomain. On the basis of the SLP procedure, only a few training models are required to compute local sensitivity of the response functions using a subdomain linear regression. SLP is a linear transformation with smoothness and differentiability, which makes it particularly compatible with Newton-like gradient-based optimization algorithms. Furthermore, we also introduce an adaptive scheme, named weighting smooth local parameterization (WSLP), in which the minimization algorithm adaptively determines the weighting coefficients and the optimal domain decomposition (DD) correspondingly, to mitigate the negative effects of an inappropriate DD strategy. We support our framework with numerical experiments for a four-variable toy model and a modified version of the sensitivity analysis of the impact of geological uncertainties on production (SAIGUP) model with spatially dependent parameters. Comparisons with previous GLR-DGN have shown that our new framework can generate comparable and even better results with significantly reduced computational cost. This SLP has high scalability, because the number of training models depends primarily on the number of local parameters in each subdomain and not on the dimension of the underlying full-order model. Activating more subdomains results in fewer local parameter patterns and enables us to run fewer training models. For a large-scale case study in this work, to optimize 412 global parameters, SLR-DGN needs only 100 initial-model simulations. In comparison to GLR-DGN where the parameters are defined over the entire domain, the central-processing-unit cost is reduced by a factor of several orders of magnitude, while retaining reasonable accuracy.

Similarity Preserving Deep Asymmetric Quantization for Image Retrieval

Quantization has been widely adopted for large-scale multimedia retrieval due to its effectiveness of coding highdimensional data. Deep quantization models have been demonstrated to achieve the state-of-the-art retrieval accuracy. However, training the deep models given a large-scale database is highly time-consuming as a large amount of parameters are involved. Existing deep quantization methods often sample only a subset from the database for training, which may end up with unsatisfactory retrieval performance as a large portion of label information is discarded. To alleviate this problem, we propose a novel model called Similarity Preserving Deep Asymmetric Quantization (SPDAQ) which can directly learn the compact binary codes and quantization codebooks for all the items in the database efficiently. To do that, SPDAQ makes use of an image subset as well as the label information of all the database items so the image subset items and the database items are mapped to two different but correlated distributions, where the label similarity can be well preserved. An efficient optimization algorithm is proposed for the learning. Extensive experiments conducted on four widely-used benchmark datasets demonstrate the superiority of our proposed SPDAQ model.

A New Learning Automata-Based Pruning Method to Train Deep Neural Networks

Deep neural network are one of the most powerful model for machine learning, which can learn the underlying patterns automatically from a large amount of data. So it can be extensively used in more and more Internet-of-Things (IoT) applications. However, the training of deep models is difficult, suffering from overfitting and gradient vanishing problem. Besides, the large amount of parameters and multiplication operations make it impractical for most deep learning models to directly execute on target hardware. In this paper, we propose a method of gradually pruning the weakly connected weights to improve the traditional stochastic gradient descent. And we adopt a reinforcement learning method called learning automata to find the weakly connected weights on account of its strong policy-making ability in stochastic and nonstationary environment. Our proposed method can learn a more effective and sparsely connected architecture during training from the initially fully connected neural networks. The experiments on MNIST show that our method have stronger power to defeat overfitting and can get better generalization performance on test set. Meanwhile, the thin and sparsely connected model we get can be more suitable for IoT applications.

Initial state perturbations as a validation method for data-driven fuzzy models of cellular networks

BackgroundData-driven methods that automatically learn relations between attributes from given data are a popular tool for building mathematical models in computational biology. Since measurements are prone to errors, approaches dealing with uncertain data are especially suitable for this task. Fuzzy models are one such approach, but they contain a large amount of parameters and are thus susceptible to over-fitting. Validation methods that help detect over-fitting are therefore needed to eliminate inaccurate models.ResultsWe propose a method to enlarge the validation datasets on which a fuzzy dynamic model of a cellular network can be tested. We apply our method to two data-driven dynamic models of the MAPK signalling pathway and two models of the mammalian circadian clock. We show that random initial state perturbations can drastically increase the mean error of predictions of an inaccurate computational model, while keeping errors of predictions of accurate models small.ConclusionsWith the improvement of validation methods, fuzzy models are becoming more accurate and are thus likely to gain new applications. This field of research is promising not only because fuzzy models can cope with uncertainty, but also because their run time is short compared to conventional modelling methods that are nowadays used in systems biology.

A Three-Stage Accelerometer Self-Calibration Technique for Space-Stable Inertial Navigation Systems.

As a specific force sensor, the tri-axis accelerometer is one of the core instruments in an inertial navigation system (INS). During navigation, its measurement error directly induces constant or alternating navigation errors of the same order of magnitude. Moreover, it also affects the estimation accuracy of gyro drift coefficients during the initial alignment and calibration, which will indirectly result in navigation errors accumulating over time. Calibration can effectively improve measurement accuracy of the accelerometer. Device-level calibration can identify all of the parameters in the error model, and the system-level calibration can accurately estimate part of these parameters. Combining the advantages of both the methods and making full use of the precise angulation of the space-stabilized platform, this paper proposes a three-stage accelerometer self-calibration technique that can be implemented directly in the space-stable INS. The device-level calibration is divided into two steps considering the large amount of parameters. The first step is coarse calibration, which identifies parameters except for the nonlinear terms, and the second step is fine calibration, which not only identifies the nonlinear parameters, but also improves the accuracy of the parameters identified in the first step. The follow-on system-level calibration is carried out on part of the parameters using specific force error and attitude error to further improve the calibration accuracy. Simulation result shows that by using the proposed three-stage calibration technique in the space-stable INS, the estimation accuracy of accelerometer error can reach order of magnitude. Experiment results show that after the three-stage calibration, the accuracy of latitude, longitude, and attitude angles has increased by over 45% and the accuracy of velocity has increased by over 22% during navigation.

A method of organizing data for storing the results of tests run on electrical products

When testing electrical products, a large number of different parameters are analyzed, processed, and stored. To ensure long-term storage and fast processing of the data, the way in which time series are stored needs to be optimized. Existing solutions do not allow simultaneously collecting a large amount of parameters with different structure and high accuracy. A data storage method based on isolation and storage of the approximating function and errors is suggested. An expression that determines the efficiency of the approximation by computing the Shannon information entropy is obtained. This article considers the effect produced by the accuracy of source data storage upon the value of entropy and hence the final volume of the data. Permissible errors of the change in many parameters of electrical products are within the range of 0.5–1.0%, which enables one use to storage of approximating functions instead of the numerical source data. To approximate the parameter values, an intermediate storage—or a buffer—is required, for which time series’ databases could be used. The suggested method can be easily integrated in databases of document-oriented type. The flexible structure of representation of testing parameter values, along with schemaless representation of document- oriented databases, allow construction and integration of information systems of varying complexity.

Application of Large-Scale Inversion Algorithms to Hydraulic Tomography in an Alluvial Aquifer.

Large-scale inversion methods have been recently developed and permitted now to considerably reduce the computation time and memory needed for inversions of models with a large amount of parameters and data. In this work, we have applied a deterministic geostatistical inversion algorithm to a hydraulic tomography investigation conducted in an experimental field site situated within an alluvial aquifer in Southern France. This application aims to achieve a 2-D large-scale modeling of the spatial transmissivity distribution of the site. The inversion algorithm uses a quasi-Newton iterative process based on a Bayesian approach. We compared the results obtained by using three different methodologies for sensitivity analysis: an adjoint-state method, a finite-difference method, and a principal component geostatistical approach (PCGA). The PCGA is a large-scale adapted method which was developed for inversions with a large number of parameters by using an approximation of the covariance matrix, and by avoiding the calculation of the full Jacobian sensitivity matrix. We reconstructed high-resolution transmissivity fields (composed of up to 25,600 cells) which generated good correlations between the measured and computed hydraulic heads. In particular, we show that, by combining the PCGA inversion method and the hydraulic tomography method, we are able to substantially reduce the computation time of the inversions, while still producing high-quality inversion results as those obtained from the other sensitivity analysis methodologies.

A Novel Design of Energy-saving Wireless Intelligent Agricultural Detecting System

In order to meet the needs of modern intelligent agricultural detection, this paper designs and develops a novel energy-saving intelligent wireless agricultural detecting system combined with technologies green power supply, wireless transmission and digital control. This system can be powered by green energy resources such as solar energy or wind power and combined with digital control and wireless communication technologies it can achieve collection of a large amount of parameters in terms of modern agricultural cultivation, which will meet the needs of different application scenarios. Results of repeated experiments have shown that it has such advantages as being green and environment-friendly, flexible and reliable.

A Parallel Genetic Algorithm for Configuring Defect Detection Methods

The detection of defects in steel strips is a very important task which can improve the performance of factories by giving the possibility of early and real-time detection of defects. Defect detection methods have such a large amount of parameters that makes finding the best configuration a complex task. The search space of the value of these parameters is pretty large also, so it is necessary to use a search algorithm in order to reduce the computing time. In this article a genetic algorithm is developed for solving this search problem. The genetic algorithm looks for an optimal or sub-optimal solution without examining the whole search space. In addition, the computing time can be reduced by running the algorithm on a grid of computers. The genetic algorithm designed allows a near-optimal configuration of defect detection methods in a short time.

Developing a Three-Dimensional Geometric Framework for Greening Buildings’ Façade

It is considerably challenging to alter the design of existing large and tall buildings to make them rely on natural energy due to many reasons. Among them is the large amount of parameters and geometric measurements required, such as height, width, plot ratio, aspect ratio, windows dimensions, etc. In this research, a new automatic algorithm is developed to model the existing buildings façade and extract several important parameters for building greening. The main contributions of this research are the automatic analysis of the digital building model and the detailed microclimatic analysis for each feature within the building façade. The developed algorithm starts with an automatic digital three-dimensional modeling of the building façade, followed by an automated extraction of the required parameters to alter the façade design such as orientation, height, and width. Results show that the proposed method offers very high accuracy and time/cost effective way for parametric modeling of the existing buildings’ façade.

Real‐time direct integration of reduced solid dynamics equations

SUMMARYA new method is developed here for the real‐time integration of the equations of solid dynamics based on the use of proper orthogonal decomposition (POD)–proper generalized decomposition (PGD) approaches and direct time integration. The method is based upon the formulation of solid dynamics equations as a parametric problem, depending on their initial conditions. A sort of black‐box integrator that takes the resulting displacement field of the current time step as input and (via POD) provides the result for the subsequent time step at feedback rates on the order of 1 kHz is obtained. To avoid the so‐called curse of dimensionality produced by the large amount of parameters in the formulation (one per degree of freedom of the full model), a combined POD–PGD strategy is implemented. Examples that show the promising results of this technique are included. Copyright © 2014 John Wiley & Sons, Ltd.

Parameter Estimation of Dynamic Load Model in Power System by using Measured Data

In this paper, we propose a dynamic load model considering change of flux inside induction motor for transient stability analysis and we present a new method for estimating unknown parameters of the dynamic load model. The model is a parallel composite of a constant impedance load and an induction motor load behind a series constant reactance. An adequate dynamic load model is essential for evaluating power system stability and this model can represent the behavior of actual load by using appropriate parameters. However, the problem of this model has a large amount of parameters and it is not easy to estimate a lot of unknown parameters. Therefore we propose an estimating method based on Particle Swarm Optimization (PSO), which is a non-linear optimization method, by using measured instantaneous voltage sag data. We estimated parameters of the actual load by PSO to minimize error between calculated and measured power of load after a load drop due to voltage sag by the accident. It was confirmed that the proposed method was successful and PSO is effective in parameter estimation.