The Differential Evolution-Support Vector Regression (DE-SVR) algorithm is designed to model the small-signal intrinsic noise behavior of GaN HEMT. It not only overcomes the local minimization shortcoming of the Back ...
详细信息
The Differential Evolution-Support Vector Regression (DE-SVR) algorithm is designed to model the small-signal intrinsic noise behavior of GaN HEMT. It not only overcomes the local minimization shortcoming of the Back Propagation (bp) algorithm, but also uses the DE (Differential Evolution) algorithm to obtain the best parameter c (punishment factor) and parameter g (variance of kernel function) of the Support Vector Regression (SVR) algorithm. In order to validate the superiority of the DE-SVR algorithm, the experiment compares the modeling effects of bp algorithm, SVR algorithm, and DE-SVR algorithm in modeling the small-signal intrinsic noise model of GaN HEMT. The experimental results show that there are obvious advantages for the DE-SVR algorithm in modeling the small-signal intrinsic noise characteristics of GaN HEMT.
This paper presents a neuro-based approach for annual transport energy demand forecasting by several socio-economic indicators. In order to analyze the influence of economic and social indicators on the transport ener...
详细信息
This paper presents a neuro-based approach for annual transport energy demand forecasting by several socio-economic indicators. In order to analyze the influence of economic and social indicators on the transport energy demand, gross domestic product (GDP), population and total number of vehicles are selected. This approach is structured as a hierarchical artificial neural networks (ANNs) model based on the supervised multi-layer perceptron (MLP), trained with the back-propagation (bp) algorithm. This hierarchical ANNs model is designed properly. The input variables are transport energy demand in the last year, GDP, population and total number of vehicles. The output variable is the energy demand of the transportation sector in Million Barrels Oil Equivalent (MBOE). This paper proposes a hierarchical artificial neural network by which the inputs to the ending level are obtained as outputs of the starting levels. Actual data of Iran from 1968-2007 is used to train the hierarchical ANNs and to illustrate capability of the approach in this regard. Comparison of the model predictions with conventional regression model predictions shows its superiority. Furthermore, the transport energy demand of Iran for the period of 2008 to 2020 is estimated.
By analyzing bp training model, a solution named division and assembly by deducing the size of bp neural network to overcome entering the local best points is introduced. The dividing process is realized by dividing a...
详细信息
By analyzing bp training model, a solution named division and assembly by deducing the size of bp neural network to overcome entering the local best points is introduced. The dividing process is realized by dividing a big bp neural network into several small ones, each of which can carry out study alone. Thereafter, they can be assembled to become the quondam big bp neural work. This approach is analyzed in detail with an example.
In this paper, the Multi-Parameters Experimental Method is adopted in the plain cascade experiment. The improved bp network is used for experiment data analyzing. It is found that the characteristics of airfoil bounda...
详细信息
In this paper, the Multi-Parameters Experimental Method is adopted in the plain cascade experiment. The improved bp network is used for experiment data analyzing. It is found that the characteristics of airfoil boundary are determined by re-attached vortex on airfoil surface at low Re. And the airfoil reverse flow is the same as the thin airfoil flow, in which the stall working condition could be reached at relatively low attack angle.
This article analyses the bp neural network studying and execution algorithm in a single computer, then proposes and uses PVM construct neural network in more computers. The implementation of parallel neural network i...
详细信息
This article analyses the bp neural network studying and execution algorithm in a single computer, then proposes and uses PVM construct neural network in more computers. The implementation of parallel neural network is flexible while being applied to deal with the high dependability and extensive data problems. Besides, the designs and implementation of the bp parallel neural network can use the model to other neural networks.
Synthetic aperture radar (SAR) can detect moving targets on the ground/sea, and high-resolution imaging on the ground/sea has critical applications in both military and civilian fields. This paper attempts to use a sp...
详细信息
Synthetic aperture radar (SAR) can detect moving targets on the ground/sea, and high-resolution imaging on the ground/sea has critical applications in both military and civilian fields. This paper attempts to use a spaceborne SAR system to detect and image moving targets in the air for the first time. Due to the high velocity of aerial targets, they usually appear as two-dimensional range and azimuth direction defocus in SAR images, and clutter will also have a profound impact on target detection. To solve the above problems, a method of detecting and focusing on a spaceborne SAR target based on a two-dimensional velocity search is proposed by combining the bp algorithm. According to the current environment of the aerial target and the number of system channels, the clutter suppression methods are set and combined with two-dimensional velocity search with different precision, the Shannon entropy under different search velocity groups is used to obtain the search velocity group closest to the actual velocity and realize the integrated processing of moving target detection-focused imaging parameter estimation. Combined with simulation data, the effectiveness of the proposed method is verified.
A fast learning algorithm for training multilayer feedforward neural networks (FNN's) by using a fading memory extended Kalman filter (FMEKF) is presented first, along with a technique using a self-adjusting time-...
详细信息
A fast learning algorithm for training multilayer feedforward neural networks (FNN's) by using a fading memory extended Kalman filter (FMEKF) is presented first, along with a technique using a self-adjusting time-varying forgetting factor. Then a U-D factorization-based FMEKF is proposed to further improve the learning rate and accuracy of the FNN, In comparison with the backpropagation (bp) and existing EKF-based learning algorithms, the proposed U-D factorization-based FMEKF algorithm provides much more accurate learning results, : using fewer hidden nodes. It has improved convergence rate and numerical stability (robustness). In addition, it is less sensitive to start-up parameters (e.g., initial weights and covariance matrix) and the randomness in the observed data. It also has good generalization ability and needs less training time to achieve a specified learning accuracy. Simulation results in modeling and identification of nonlinear dynamic systems are given to show the effectiveness and efficiency of the proposed algorithm.
When remote sports injuries are traditionally monitored, the feedback is not timely, and there are some problems such as low running speed and poor accuracy of sports injury monitoring. Therefore, this paper designs a...
详细信息
When remote sports injuries are traditionally monitored, the feedback is not timely, and there are some problems such as low running speed and poor accuracy of sports injury monitoring. Therefore, this paper designs a remote sports injury monitoring method based on a wireless sensor network. Firstly, the terminal node of the sports injury monitoring process is designed, and three-terminal devices are tied to the experimental object's body to collect motion information, to realize the collection of human motion information based on ZigBee wireless sensor network;Secondly, the USB module circuit interface is designed to realize the series connection of each line, and the local processing ability of network nodes is used to make a centralized decision. Then, the skeleton coordinate system is constructed, and the rotation of the human skeleton is measured by an inertial sensor. Through a variety of posture fitting, the error of remote sports injury monitoring is reduced from the two directions of joint error and muscle error. Finally, the training sample set is learned through the bp algorithm, the fitness function of the genetic algorithm is obtained, the external structural parameters of the adaptive neural network model are adjusted, the discrimination deviation and fitness function are calculated, the adaptive neural network model with the best generalization ability is output, and the local processing ability of remote sports injury monitoring method is improved combined with wireless sensor network technology, The design of remote sports injury monitoring method based on wireless sensor network is realized. The experimental results show that the accuracy of the method is 99%, the average time delay is 1 s, and the accuracy of the method is 92% even with noise. Therefore, the method can effectively improve the running speed of the remote sports injury monitoring method and improve the accuracy of sports injury monitoring.
In the Internet of Things, sensor nodes collect environmental information and utilize lossy compression for saving storage space. To achieve this objective, high-efficiency compression of the continuous source should ...
详细信息
In the Internet of Things, sensor nodes collect environmental information and utilize lossy compression for saving storage space. To achieve this objective, high-efficiency compression of the continuous source should be studied. Different from existing schemes, lossy source coding is implemented based on the duality principle in this work. Referring to the duality principle between the lossy source coding and the channel decoding, the belief propagation (bp) algorithm is introduced to realize lossy compression based on a Gaussian source. In the bp algorithm, the log-likelihood ratios (LLRs) are iterated, and their iteration paths follow the connecting relation between the check nodes and the variable nodes in the protograph low-density parity-check (P-LDPC) code. During LLR iterations, the trapping set is the main factor that influences compression performance. We propose the optimized bp algorithms to weaken the impact of trapping sets. The simulation results indicate that the optimized bp algorithms obtain better distortion-rate performance.
An energy-saving scheme for pumping units via intermission start-stop performance is proposed. Because of the complexity of the oil extraction process, Fuzzy Neural Network (FNN) intelligent control is adopted. The st...
详细信息
An energy-saving scheme for pumping units via intermission start-stop performance is proposed. Because of the complexity of the oil extraction process, Fuzzy Neural Network (FNN) intelligent control is adopted. The structure of the Takagi-Sugeno (T-S) fuzzy neural network model is introduced and modified. FNNs are trained with sample information from oil fields and expert knowledge. Finally, pumping unit energy-saving FNN software, which cuts down power costs substantially, is presented.
暂无评论