In this paper, we use a deep learning algorithm to optimize the loss function of Chinese data, dynamically adjust the learning rate by gradient descent method, and adopt an exponential moving ...weighted average to deal with language information. The second-order information of the historical gradient is processed to calculate the change value of the objective function of the Chinese language variables, and the learning rate is scaled accordingly, which is inversely proportional to the teaching reform model. Ultimately, the error between the objective function value and the predicted value is optimized with a smaller learning rate, and the neuron learning rate is scaled to achieve the best solution for the language teaching reform of ancient Chinese and modern Chinese. The accuracy rate of the deep learning algorithm is verified to be as high as 0.9, which provides reliable technical support for teaching reform. The reform’s weighting of teaching content to 0.471 underscored its significance in teaching. Fifty-three percent of the students showed strong interest in the teaching method reform, highlighting the popularity of teaching mode innovation. The teaching rating before the reform increased by 6 points compared to that after the reform, showing the positive impact of the reform on teaching.
With the popularization and development of communication technology, the resource allocation problem of satellites has become a hot research topic. This study proposes a model for satellite resource ...allocation through the development of a beam mobilization system model and problem modeling. Then, on the basis of the linear stochastic gradient descent method, the classification accuracy of the algorithm is improved by adjusting the algorithm-solving method. Then its weight assignment method is improved to get the improved weighted linear stochastic gradient descent method. Using the optimized weighted linear stochastic gradient descent method to solve the satellite resource allocation problem from the perspective of the original problem, a model based on IWLSGD is designed and tested for performance. Through simulation experiments, the beam allocation service value of the satellite resource allocation model in this paper is 193, which is 4.32% and 3.21% higher than that of LSGD and WLSGD, respectively, and iterative convergence is faster, and its operation time and service value performs the best under different numbers of communication time slots. The system revenue, system access success rate, and system satisfaction under the interference environment between LEO and GEO are maintained at 15208, 0.8~1.0, and 85%, and keep around 7500, 1, and 75% under 5G base station interference. The satellite resource allocation model in this paper can effectively improve the utilization efficiency of communication resources and better adapt to dynamically changing interference scenarios.
Agricultural pests and diseases critically impact the quality and yield of crops, thereby underscoring the practical importance of their automatic monitoring, identification, and timely management in ...agricultural production. This study develops a targeted detection model using a deep learning approach, specifically by enhancing the Faster R-CNN algorithm. Modifications were implemented in three key areas of the basic Faster R-CNN: First, the DIOU-NMS technique was employed to optimize the ancillary network during the feature extraction phase. Secondly, an attention mechanism along with an SE module was integrated within the DIOU-NMS to augment the network’s capability. During the training phase, optimization was facilitated through stochastic gradient descent. The efficacy of the refined Faster RCNN model was established via ablation studies, and its performance was benchmarked against existing methodologies for small and general target detection. Results indicate that the enhanced Faster R-CNN framework surpasses conventional small target and generic detection models in accuracy, achieving a 6.4% higher detection rate for various pest categories compared to its predecessor. The findings affirm the potential of the advanced Faster R-CNN in effective agricultural pest detection. Furthermore, this paper advocates a tripartite strategy for pest management, encompassing phytosanitary measures, agricultural interventions, and chemical controls.
In this paper, a web-based teaching platform is constructed based on the background of big data, which is modeled by using a neural network algorithm. First, the neural network structure is output, ...and the error function is minimized by dynamic iterations using the activation function as neurons. Then, the error values were trained with implicit nodes, the activation function was modeled nonlinearly, and the sample set was extracted to define the cost function. Finally, the optimal particle is trained iteratively using the gradient descent method to derive the optimal solution, thus completing the construction of a web-based teaching platform based on the big data background. The experimental results show that after using this platform for online ideological education teaching, the percentage of students visiting online courses every day is 35%, which is 29% higher than that before using it. Therefore, to improve the construction level of ideological and political education online courses, it is necessary to strengthen the construction of ideological and political education online course platform resources, improve the informatization level of ideological and political education teacher teams, and promote the integration of ideological and political education online courses and classroom teaching.
This article proposes an algorithm correcting the field orientation inaccuracy caused by resolver periodic error and rotor time constant variation. It is found that the periodic error brings harmonic ...current vectors and torque ripples while the rotor time constant variation makes dq currents deviate from the reference values. The cross-product of fundamental and harmonic current vectors contains the information of resolver periodic error. The gradient descent method with a variable step is introduced to compensate the periodic error. During the process, the cross-product is minimized. A new form of cross-product calculation is derived to reduce the computation time. The rotor time constant is identified by the reactive-power based model reference adaptive system (MRAS). Voltages and currents in the stationary reference frame are used in the reference model to avoid the influence of resolver periodic error. Appropriate parameters of the MRAS are selected to guarantee the stability. The algorithm is integrated with an induction motor drive which is based on indirect field-oriented control. The feasibility of the proposed algorithm is verified by simulation results and experiments.
Abstract
A fault diagnosis and localization approach for distributed distribution networks is created using an upgraded quantum genetic algorithm to swiftly identify and detect flaws in the network. ...In this method, the dynamic rotation strategy in gradient descent method is used to update the quantum gate to enhance the convergence speed, that is, the gradient descent quantum genetic algorithm is constructed. The results of single fault and multiple fault simulation test on the distribution network model of regional node of distributed power supply show that the average iteration of gradient descent quantum genetic algorithm 85.36, 86.35, 88.24, and 88.69 times can reach the target optimal value. In four different cases, the algorithm of gradient descent quantum genetic algorithm can reach the optimal by iterating 88, 91, 92, and 90 times, respectively. Compared with other algorithms, the convergence rate of gradient descent quantum genetic algorithm is the fastest in the four experimental cases. The consistency between the output score and the real score of the gradient descent quantum genetic algorithm is above 0.9. The results above show that the algorithm is effective. The optimization ability and stability of the algorithm are also stronger, and it has certain application potential.
The high-speed motorized spindle is the core component of high-speed and high-precision machining, and its compact structure leads to internal heat accumulation and thermal deformation. Therefore, it ...is of great significance to control the temperature rise of the motorized spindle. In order to effectively control the temperature rise of the motorized spindle, a new spiral cooling system is used to analyze the internal heat transfer mechanism of the high-speed motorized spindle, and the heat transfer coefficients of the spiral cooling system and the motorized spindle system are optimized based on the gradient descent method combined with experimental data. The optimized heat transfer coefficient is taken as the boundary condition of the finite element model, and the temperature field prediction model is established to analyze the influence of the spiral cooling system on the temperature field of the motorized spindle. Through experiments, the cooling capacity of the spiral cooling system is verified, and the optimized temperature field simulation data are compared with the experimental data to verify the feasibility of the gradient descent method in constructing the temperature field prediction model of the motorized spindle. It provides a basis for the intelligent control of the thermal performance of the motorized spindle.
Due to the slow convergence rates and reliance on global information, distributed optimization methods, such as alternating direction method of multipliers (ADMM) and distributed gradient descent ...method (DGDM), have been difficult to meet the needs of solutions for the large-scale distributed system. In this paper, a distributed multi-step gradient descent method (DMGDM) combined with the allocation of upper bounds of second derivatives, has been proposed to tackle the economic dispatch problem (EDP) of microgrid (MG). At first, the agent assigns the reciprocal of its own upper bound of the second derivative according to the product of out-degree and upper bounds of second derivatives from the neighbors, which is the key to our method. Further, the weight matrix is constructed through the negotiation among neighbors in a distributed manner. Additionally, the distributed weight matrix relaxes the convergence conditions of the proposed method so that momentum parameters can be tuned without the need for global information. Numerical examples show that the convergence rate of our method is faster than the ADMM and the DGDM. Finally, a bi-layer optimization model of the EDP of MG is built via Matlab/Simulink, and the results show that the proposed method can realize the optimal dispatch of controllable distributed generators (DGs) in the MG.
In this paper, we introduce a novel and robust approach to quantized matrix completion. First, we propose a rank minimization problem with constraints induced by quantization bounds. Next, we form an ...unconstrained optimization problem by regularizing the rank function with Huber loss. Huber loss is leveraged to control the violation from quantization bounds due to two properties: first, it is differentiable; and second, it is less sensitive to outliers than the quadratic loss. A smooth rank approximation is utilized to endorse lower rank on the genuine data matrix. Thus, an unconstrained optimization problem with differentiable objective function is obtained allowing us to advantage from gradient descent technique. Novel and firm theoretical analysis of the problem model and convergence of our algorithm to the global solution are provided. Another contribution of this letter is that our method does not require projections or initial rank estimation, unlike the state-of-the-art. In the Numerical Experiments section, the noticeable outperformance of our proposed method in learning accuracy and computational complexity compared to those of the state-of-the-art literature methods is illustrated as the main contribution.
The scheme of dynamic management of traffic and activity of message sources with different priority of service is considered. The scheme is built on the basis of the neuroprognostic analysis model ...and the gradient descent method. For prediction and early detection of overload, the apparatus of the general theory of sensitivity with indirect feedback and control of activity of message sources is used. The control algorithm is started at the bottleneck of the network node. It uses a recursive prediction approach where the neural network output is referred to as many steps as defined by a given prediction horizon. Traffic with a higher priority is served without delay using the entire available bandwidth. Low-priority traffic will use the remaining bandwidth not used by higher-priority traffic. An algorithm for estimating the maximum available bandwidth of a communication node for traffic with a low service priority has been developed. This approach makes it possible to improve the efficiency of channel use without affecting the quality of service for high-priority traffic.