12 results about "Function optimization" patented technology

This invention discloses a water treatment dosing control method and system based on quadratic programming, belonging to the field of water treatment process control and optimization technology. It collects historical water treatment operation data and constructs a mechanistic feature set, using the mechanistic feature set as the independent variable and turbidity reduction as the target variable. The turbidity reduction is used to represent the change in flocculation or sedimentation of suspended impurities in the water, and a full-variable regression model is constructed. Variables in the mechanistic feature set are screened, and the full-variable regression model is optimized using a stepwise regression method to obtain a simplified prediction model. Real-time influent water quality parameters are acquired and input into the simplified prediction model. The process of maximizing turbidity reduction in the simplified prediction model is transformed into minimizing a convex loss function, and the convex loss function is iteratively optimized using a hierarchical constrained projection gradient descent method to obtain the optimal dosing scheme. Through mechanism-driven modeling, convex function optimization, and hierarchical constrained projection, intelligent, efficient, and reliable control of the dosing process is achieved.

The application designs a natural evolution strategy optimization method based on variance reduction technology, aiming at solving the problem of large gradient estimation variance and slow convergence speed in traditional natural evolution strategy. The method combines variance reduction technology, introduces global average gradient and small batch gradient correction in the optimization process, significantly reduces the variance of gradient estimation, improves the optimization accuracy and efficiency. Specifically, the method includes the following steps: first, initialize the optimization parameters, including the initial solution, learning rate and step size; then calculate the global gradient by input data, which is used for subsequent correction; in the inner loop, calculate the objective function value and gradient for small batch perturbation samples, and combine the global gradient for variance correction to update the optimization parameters. Compared with the traditional method, the application has the advantages of high gradient estimation accuracy and fast convergence speed, and is suitable for black box objective function optimization scenarios, and has wide application value in the field of machine learning model training.

The application discloses a kind of excavator bucket tooth tip trajectory planning method with arm deflection and medium.The current excavator bucket task corresponding to target excavator is obtained and according to target excavator, the target posture information and the current posture information of bucket tooth tip are obtained by analysis, and the endpoint constraint condition corresponding to the starting point of current bucket task and the endpoint of current bucket task respectively, using quintic polynomial time parameterization method, the motion trajectory of bucket tooth tip in three-dimensional space of target excavator is time parameterized respectively, and initial trajectory function is generated;According to the constraint condition of dynamics and hydraulic mapping model, function optimization is carried out by trajectory function optimization method, and target trajectory parameter description set is obtained, to realize the control operation of excavator bucket tooth tip.The problem that the single joint position control or manual operation based on experience is used to control excavator in prior art is solved, the difficulty degree of excavator operation and trajectory tracking error are reduced, and the operation accuracy is improved.

The application discloses a kind of connecting node design method and system of fabricated modular building, including identifying fabricated module contact area and dividing connecting node group, determining the quantity and position of connecting member of each connecting node, determining the first target stiffness of each connecting node group, the second target stiffness is obtained by using point group stiffness optimization objective function optimization, construct connecting node state prediction model, obtain initial stiffness target value by reducing second target stiffness, adjusting connecting node controllable parameter according to initial stiffness target value and connecting node state prediction model, build fabricated modular building, calculate cumulative error and cumulative adaptive tolerance limit, compare data Real-time fine-tuning connecting node part controllable parameter.The method not only can improve the efficiency and accuracy of the connecting node design of fabricated modular building, but also has good explainability, which can be directly applied to the connecting node design system of fabricated modular building.

The learning apparatus for machine learning a score function to be used for two-class classification includes a score function optimization unit that updates, by setting, as an index, a total value of a value obtained by multiplying an AUC by a weight λ which is a positive real number and a value obtained by multiplying a pAUC by a weight (1−λ), a parameter of the score function through machine learning that is performed, using training data, so as to maximize the index, while decreasing the value of the weight λ. The AUC indicates an area of a region on a horizontal axis side of a curve obtained by changing a threshold for determining a positive example and a negative example in the score function. The pAUC is a value of the AUC when the value of the false positive rate is set to a specific fixed value.

The application relates to the computer technical field, and particularly provides a reduction type kernel function optimization method for a MuXiXi cloud C500, which comprises the following steps: obtaining the data length N of a to-be-executed reduction subject, and selecting a reduction subject with a data length N greater than a preset value to execute the following steps; setting the thread number in a kernel function as 1024; calculating the data number NUM to be processed by each thread; determining the processing frequency W of each thread, the vector length X, the parallel processing vector number Z and the last parallel processing vector number Z' according to the NUM; then setting the kernel function starting parameters, and sequentially performing reduction operations on the vectors in each thread, the vectors between the threads, the processing frequency of each thread and the data of all threads. According to the above technical scheme, for a reduction subject with a large amount of input data, the optimization of the configuration parameters is realized in 1024 threads in one block according to the data length of the reduction subject, and the kernel function development efficiency is improved.

The application provides a database intelligent index recommendation method and system based on reinforcement learning, which comprises the following modules: (1) an initialization module; (2) an action space definition module; (3) a reward function optimization module, which selects appropriate index operations to minimize the reward function value; (4) a model training module; and (5) a deployment and optimization module. The application introduces a reward function calculation method that comprehensively considers multiple key factors such as the optimization effect and size of the index, and the reward function can more comprehensively evaluate the good and bad of an action (index operation). Compared with a single index, this multi-factor comprehensive consideration method is more in line with the complexity of the actual database performance, and is helpful to improve the accuracy and applicability of the index recommendation algorithm. The introduction of the evaluation of the index size avoids excessive consumption of system resources such as memory and disk space. Considering the resource consumption factors such as the index size in the reward function helps to avoid performance problems caused by excessively large or small index configurations, and improves the robustness and practicality of the system.

The application relates to a TDOA / AOA data supplemented electromagnetic spectrum map construction method. The method comprises the following steps: a measurement point collects a received signal to determine the received signal strength of the measurement point to obtain TDOA measurement data and AOA measurement data, a probability model is constructed to probabilistically update an electromagnetic spectrum map of a target region, a posterior probability distribution of a signal source position is obtained, a signal source position corresponding to a maximum value of the posterior probability distribution is taken as a candidate signal source position, and the received signal strength of the candidate signal source position and the received signal strength at other non-measurement point positions are estimated; a candidate point set is randomly generated in the target region, a candidate point with the maximum target function value is selected as a target point and is added to an accurate data set; after adaptive mixed kernel function optimization parameters are constructed, a Bayesian Kriging method is used to interpolate and output an electromagnetic spectrum map of the target region after supplement, and the electromagnetic spectrum map construction precision is improved.

This invention proposes a multi-objective optimization acceleration method for wind turbine controller parameters using a deep fully connected layer. This method is applied to multi-objective problems, and during the iteration process, the search agent's position is updated in three directions: proportion, exploration, and development. For situations where the external archive set is crowded, an adaptive grid strategy is adopted to update non-dominated solutions to address the congestion. A roulette wheel strategy is used to select the globally optimal position. A deep fully connected layer model is added during the method iteration process to improve convergence speed. This multi-objective optimization acceleration method for wind turbine controller parameters can solve the bi-objective optimization problem of doubly-fed induction generator (DFIG) wind turbine controller parameters, enabling rapid acquisition of optimal solution sets from multiple schemes. It optimizes the diversity and convergence of the solution set in multi-objective methods, reduces the required optimization time, and improves the running speed of the optimization method.

This invention relates to the field of logistics distribution route planning, and more particularly to a low-carbon logistics distribution route planning method based on multi-objective optimization. The method includes: constructing a directed graph based on the logistics distribution network and obtaining the travel distance, travel time, and transportation economic cost of each route segment; calculating the carbon emissions of each route segment based on its travel distance and travel time; constructing a multi-objective function optimization model based on the travel time, carbon emissions, and transportation economic cost of each route segment; and using an improved non-dominated sorting genetic algorithm based on the multi-objective function optimization model to generate preliminary distribution route schemes, and introducing a service time window penalty mechanism to identify locally optimized candidate route segments and perform route optimization to obtain the optimal distribution route scheme. This method solves the problems of traditional logistics distribution route planning methods, such as insufficient integration of actual traffic conditions, unreasonable carbon emission structure design, lack of customer service time window constraints, and inadequate objective optimization strategies.

The application discloses an agent reinforcement learning reward function optimization method based on VLM closed-loop feedback. First, an initial reward function code and a parameter configuration file are generated according to natural language task instructions by using a large language model; then, the reward function is used to drive a robot arm to perform reinforcement learning strategy training, and a video stream of task execution is recorded and the task success rate is calculated. When the preset failure trigger condition is met, the parameter update is paused. An algorithm based on perceptual hashing and time sequence hybrid weight is used to extract a key frame sequence from the video, a visual language model is used for semantic diagnosis of the key frame, and improvement suggestion is fed back to the large language model, and the reward function is optimized and hot reloaded into the training process. By introducing the visual language model to construct an automatic feedback correction closed loop, the problem that the large language model generates a reward function lacking physical verification and leading to failure is solved, and the training efficiency and success rate of the agent for complex operation tasks are significantly improved.

This invention discloses a novel swarm intelligence optimization algorithm integrating multiple strategies. The method steps are as follows: S1, initialize the Tibetan fox population and dynamic territory; S2, adaptively adjust the territory division using nonlinear dynamic parameters combined with population distribution; S3, integrate PSO, DE, and gradient search strategies to dynamically adjust the selection probability to achieve precise local development; S4, design an adaptive migration mechanism; S5, maintain population diversity through subpopulation co-evolution and periodic information exchange; S6, add a globally optimal guided pattern search to improve convergence accuracy; S7, complete population update using elite retention and dynamic elimination strategies. Compared with existing technologies, this invention's core parameters are dynamically adjusted nonlinearly, balancing exploration and development. Experiments verify that it has better accuracy and robustness in multi-function optimization, and its performance in practical applications is significantly better than PSO and GA. It can be widely used in engineering optimization, machine learning hyperparameter tuning, and other fields.