A sensor algorithm parameter tuning method based on non-convex optimization and entropy weight method

Abstract

The application discloses a sensor algorithm parameter tuning method based on non-convex optimization and entropy weight method, which comprises the following steps: S1, setting an optimization objective function of a sensor, and initializing sensor parameters and data collection; S2, optimizing the optimization objective function of the sensor by using a stochastic gradient descent method; S3, introducing a SiLU activation function to enhance the performance of the optimization process; S4, after completing local parameter tuning, introducing an entropy weight method to dynamically adjust the weight of the sensor node; and S5, combining a centralized and distributed global data synchronization mechanism to ensure that the tuning result is shared in a large-scale sensor network. The sensor algorithm parameter tuning method based on non-convex optimization and entropy weight method aims to solve the global optimization problem of sensor parameter tuning in a large-scale and complex environment. The traditional tuning algorithm in a large-scale sensor network is limited to local optimization.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of sensor parameter optimization, and particularly relates to a sensor algorithm parameter optimization method based on non-convex optimization and entropy weight method. BACKGROUND

[0002] In large-scale sensor networks, sensor parameter optimization is crucial for ensuring data collection accuracy and system stability. Traditional optimization methods are mostly based on classic optimization algorithms such as particle swarm optimization (PSO) and genetic algorithm (GA). They perform well in local parameter adjustment but struggle with global optimization in high-dimensional spaces and complex environments, often falling into local optima. PSO, while showing good global search capabilities in many optimization problems, often gets stuck in local optima when dealing with high-dimensional non-convex optimization problems, especially in complex, multi-modal non-convex spaces. PSO relies on the cooperation between individuals and groups, which can lead to stable search directions and loss of global optimization exploration ability. GA, as an evolutionary algorithm, explores the solution space through selection, crossover, and mutation operations. However, GA often requires longer iteration periods to converge in high-dimensional complex environments and is easily affected by local solutions in non-convex spaces, resulting in low search efficiency.

[0003] Furthermore, existing methods struggle with non-linear, non-convex optimization problems when dealing with large-scale sensor networks. Therefore, a new optimization method is needed to effectively handle large-scale, multi-dimensional sensor network optimization in complex environments. SUMMARY

[0004] The main purpose of the present application is to provide a sensor algorithm parameter optimization method based on non-convex optimization and entropy weight method, aiming to solve the global optimization problem of sensor parameter optimization in large-scale, complex environments. To overcome the limitations of traditional optimization algorithms in large-scale sensor networks, which are confined to local optima, the method optimizes the architecture and data synchronization mechanism to improve the overall network optimization efficiency and accuracy. The method mainly uses stochastic gradient descent (SGD) and non-convex projection gradient method to solve non-convex optimization problems and introduces SiLU (Sigmoid-Weighted Linear Unit) activation function to improve the non-linear performance and stability of the optimization process. At the same time, the entropy weight method dynamically adjusts the weight of the sensor nodes, thereby improving the system's adaptability and global optimization ability.

[0005] To achieve the above purpose, the present application provides a sensor algorithm parameter optimization method based on non-convex optimization and entropy weight method, comprising the following steps:

[0006] Step S1: Set the optimization objective function of the sensor, and initialize the sensor parameters and data collection;

[0007] Step S2: (To prevent falling into local optimal solution) Optimize the optimization objective function of the sensor by random gradient descent method;

[0008] Step S3: Introduce SiLU activation function to enhance the performance of the optimization process;

[0009] Step S4: After completing the local parameter tuning, introduce entropy weight method to dynamically adjust the weight of the sensor node;

[0010] Step S5: By combining the centralized and distributed global data synchronization mechanism, ensure that the tuning results are shared in large-scale sensor networks.

[0011] As a further preferred technical solution of the above technical solution, for step S1:

[0012] Assume that the number collected by the sensor is x i , the sensor parameter is θ i , and the optimization objective is to minimize the non-convex objective function f(θ), which is in the form of:

[0013]

[0014] Where w i represents the weight of the i-th sensor, g(θ i , x i ) is a non-convex function representing the relationship between sensor performance and its parameters and data.

[0015] As a further preferred technical solution of the above technical solution, for step S2:

[0016] First, randomly initialize the parameters (θ0), then update the parameters of the sensor by calculating the gradient of the objective function, specifically:

[0017]

[0018] Where η is the learning rate, which controls the step size of parameter update; after updating the parameters each time, the parameters are projected to ensure that the parameters are within the feasible region, specifically:

[0019]

[0020] Where Proj C () is the projection function, C is the feasible region of the constraint condition, which is related to the physical or application constraints of the sensor (such as energy consumption limit or upper and lower limit of the parameter).

[0021] As a further preferred technical solution of the above technical solution, for step S3:

[0022] The formula of the SiLU activation function is:

[0023]

[0024] The SiLU activation function improves the smoothness and stability of the gradient update process through nonlinear processing, and can avoid the problem of gradient vanishing or explosion. In the gradient descent process, the SiLU activation function is applied to the gradient update formula to optimize the nonlinear performance of the tuning, and the updated gradient formula is:

[0025]

[0026] Through this SiLU activation function processing of the gradient update, it effectively avoids falling into local optimum in non-convex optimization, and improves the convergence speed of global optimum.

[0027] As a further preferred technical solution of the above technical solution, for step S4, the entropy weight method judges the importance (in the system) by calculating the change rate of sensor data, and dynamically adjusts the weight of each node according to the size of information entropy. The specific calculation method is: for each sensor node, the data change rate p ij Calculate the information entropy H i :

[0028]

[0029] Where p ij represents the data change rate of the i-th sensor at the j-th moment, and the higher the information entropy H i , the greater the change of the node data, and the higher priority is given in the tuning; therefore, the dynamic adjustment of the weight of the node is:

[0030]

[0031] Through the dynamic adjustment mechanism, the weight is adaptively allocated in real time, so that the nodes with large changes respond more quickly to environmental changes, thereby improving the tuning efficiency of the entire network.

[0032] As a further preferred technical solution of the above technical solution, for step S5:

[0033] (In order to ensure that the tuning results can be shared in a large-scale sensor network, the application designs a hierarchical data synchronization mechanism based on a central node) First, the central node collects information including delay and load of all nodes in the network to dynamically adjust the data synchronization frequency, flexibly set the synchronization interval time through an adaptive control algorithm, thereby responding in time when the data flow changes (both avoiding too frequent synchronization leading to network congestion and reducing the data lag problem caused by low-frequency synchronization), so that the dynamic frequency control method can realize efficient synchronization of the whole network according to the node size and data flow;

[0034] In the data transmission process, in order to guarantee the integrity of the synchronization data, a packet loss detection and compensation mechanism is introduced, the central node monitors the transmission order and loss of data packets, and once the packet loss is detected, a compensation request is triggered in time to require the corresponding node to resend the lost data packet, so as to ensure complete data transmission, in order to reduce the influence of packet loss and delay on the stability of the system, a delay processing algorithm is adopted to dynamically adjust the delay of the data packet, and the delayed data is processed through priority sorting to ensure the timeliness of the key information;

[0035] The synchronization mechanism combines the centralized and distributed schemes: the central node is responsible for the synchronization control of the whole network and the collection of key data, while the distributed nodes can also directly transmit local data to reduce the dependence on the central node and improve the response speed and stability of the synchronization (the combination of centralized and distributed data synchronization methods helps to realize stable, bidirectional data sharing and efficient data synchronization in a large-scale sensor network); after each local parameter tuning is completed, each sensor node uploads the tuning result to the central node through wireless communication, the central node collects the tuning results of all nodes and calculates the global optimal parameter global , the formula is:

[0036]

[0037] The central node feeds back the global optimal parameter to each sensor node to ensure that the parameter is optimized in the global range (this distributed data synchronization mechanism not only can greatly reduce the computing pressure of the central node, but also can improve the tuning efficiency of the whole sensor network, especially in a large-scale sensor network).

[0038] The application has the advantages of:

[0039] The application can effectively solve the non-convex optimization problem of a complex sensor network and avoid local optimal solution. By combining the non-convex optimization algorithm, the stochastic gradient descent, the projection method and the SiLU activation function, the optimization process is smoother and the effect is better. The entropy weight method realizes dynamic weight adjustment, enhances the system adaptability and efficiency. The distributed data synchronization mechanism ensures global parameter sharing, is stable in a large-scale network and is suitable for complex scenes such as smart cities. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 is a flowchart of a sensor algorithm parameter tuning method based on non-convex optimization and entropy weight method. DETAILED DESCRIPTION

[0041] The following description is used to disclose the application so that those skilled in the art can implement the application. The preferred embodiments in the following description are only used as examples, and other obvious modifications can be thought of by those skilled in the art. The basic principles of the application defined in the following description can be applied to other embodiments, modifications, improvements, equivalents and other technical solutions without departing from the spirit and scope of the application.

[0042] In the preferred embodiments of the application, those skilled in the art should note that the sensors and the like involved in the application can be regarded as prior art.

[0043] Preferred embodiments.

[0044] The application discloses a sensor algorithm parameter tuning method based on non-convex optimization and entropy weight method, comprising the following steps:

[0045] Step S1: setting an optimization objective function of a sensor and initializing sensor parameters and data collection;

[0046] Step S2: (in order to prevent falling into a local optimal solution) optimizing the optimization objective function of the sensor by a stochastic gradient descent method;

[0047] Step S3: introducing a SiLU activation function to enhance the performance of the optimization process;

[0048] Step S4: after completing local parameter tuning, introducing an entropy weight method to dynamically adjust the weight of the sensor node;

[0049] Step S5: by combining the centralized and distributed global data synchronization mechanism, ensuring that the tuning result is shared in a large-scale sensor network.

[0050] Specifically, for step S1:

[0051] Suppose that the number collected by the sensor is x i , and the sensor parameter is θ i, the optimization objective is to minimize a non-convex objective function f(θ), which is in the form of:

[0052]

[0053] where w i is the weight of the i-th sensor, g(θ i , x i ) is a non-convex function representing the relationship between sensor performance and its parameters and data.

[0054] More specifically, for step S2:

[0055] First, the parameters are randomly initialized (θ0), and then the parameters of the sensor are updated by calculating the gradient of the objective function , which is specifically:

[0056]

[0057] where η is the learning rate, which controls the step size of parameter update; after updating the parameters each time, a projection operation is performed on the parameters to ensure that the parameters are within the feasible region, which is specifically:

[0058]

[0059] where Proj C () is the projection function, and C is the feasible region of the constraint, which is related to the physical or application constraints of the sensor (such as energy consumption constraints or upper and lower limits of parameters).

[0060] Further, for step S3:

[0061] The formula of the SiLU activation function is:

[0062]

[0063] The SiLU activation function improves the smoothness and stability of the gradient update process through nonlinear processing, and can avoid the problem of gradient vanishing or explosion) In the gradient descent process, the SiLU activation function is applied to the gradient update formula to optimize the nonlinear performance of the tuning, and the updated gradient formula is:

[0064]

[0065] Through this SiLU activation function processing of the gradient update, it effectively avoids falling into local optimum in non-convex optimization, and improves the convergence speed of global optimum.

[0066] Further, for step S4, the entropy weight method judges the importance by calculating the change rate of the sensor data, and dynamically adjusts the weight of each node according to the size of the information entropy, and the specific calculation method is: for each sensor node, the data change rate p ij Calculate the information entropy H i :

[0067]

[0068] Where, p ij represents the data change rate of the i-th sensor at the j-th moment, the information entropy H i is higher, indicating that the node data changes more, and is given higher priority in the tuning; therefore, the dynamic adjustment of the weight of the node is:

[0069]

[0070] Through the dynamic adjustment mechanism, the weight is adaptively allocated in real time, so that the nodes with large changes respond more quickly to environmental changes, thereby improving the tuning efficiency of the entire network.

[0071] Preferably, for step S5:

[0072] (To ensure that the tuning result can be shared in a large-scale sensor network, the application designs a hierarchical data synchronization mechanism based on a central node) First, the central node collects all the nodes in the network, including delay and load (etc.) information, to dynamically adjust the data synchronization frequency, and flexibly sets the synchronization interval time through an adaptive control algorithm, so as to respond in time when the data flow changes (both to avoid too frequent synchronization leading to network congestion, and to reduce the problem of data lag caused by low-frequency synchronization), so that this dynamic frequency control method can realize efficient synchronization of the entire network according to the node size and data flow;

[0073] In the data transmission process, in order to guarantee the integrity of the synchronized data, a packet loss detection and compensation mechanism is introduced, the central node monitors the transmission order and loss of data packets, and once packet loss is detected, a compensation request is triggered immediately, requiring the corresponding node to resend the lost data packet, to ensure complete data transmission. In order to reduce the impact of packet loss and delay on system stability, a delay processing algorithm is used to dynamically adjust the delay of data packets, and the delayed data is processed through priority sorting to ensure the timeliness of critical information.

[0074] The synchronization mechanism combines the centralized and distributed schemes: the central node is responsible for the synchronization control of the whole network and the aggregation of key data, while the distributed nodes can also directly transmit local data to reduce the dependence on the central node and improve the response speed and stability of the synchronization (the centralized and distributed combined data synchronization method helps to realize stable, bidirectional data sharing and efficient data synchronization in a large-scale sensor network); after each local parameter optimization is completed, each sensor node uploads the optimization result to the central node through wireless communication, and the central node aggregates the optimization results of all nodes and calculates the global optimal parameter global , the formula is:

[0075]

[0076] The central node feeds back the global optimal parameter to each sensor node to ensure that the parameter is cooperatively optimized in the global range (this distributed data synchronization mechanism can not only greatly reduce the computing pressure of the central node, but also improve the optimization efficiency of the whole sensor network, especially in a large-scale sensor network).

[0077] For the present application, first, the random gradient descent method is used to process large-scale data, and a small amount of random samples is selected to calculate the gradient in each iteration, thereby improving the calculation efficiency. In the implementation process of the random gradient descent method, the specific steps include initializing the learning rate and randomly selecting the sample set. In each iteration, the method randomly selects a subset from the data to calculate the gradient, and uses the adaptive learning rate to adjust the update step of the algorithm. In order to avoid falling into local optimum, SGD usually adopts the momentum strategy, that is, the update direction of the previous times is superimposed in the update process, thereby helping to search out of the local area. In addition, when facing noisy data, SGD reduces the learning rate step by step to ensure convergence, and regularizes the data when necessary, thereby further improving the robustness of the optimization process.

[0078] The updated solution is constrained in the feasible region by using the non-convex projection gradient method, ensuring the validity of the solution. At the same time, the entropy weight method dynamically adjusts the node weight according to the data information entropy of each sensor node, so that the node with more information has greater influence in the system. First, the data fluctuation of each node in a period of time is calculated by analyzing the change rate of sensor data, so as to obtain the information entropy of each node. The weight adjustment process dynamically adjusts the node weight according to the information entropy value. The higher the entropy value, the more information the node contains, and the weight increases accordingly; otherwise, it decreases. In this way, the system can quickly respond to environmental changes and improve the adaptability. Through this dynamic adjustment, the entropy weight method effectively avoids the problem of insufficient sensitivity to environmental changes of the traditional fixed weight method, and enhances the stability of the system in a variable environment. By calculating the weight of the node in each optimization process and adjusting it in time, the system can maintain adaptability and global optimization effect in a variable environment, so as to achieve higher accuracy and robustness in positioning and navigation.

[0079] It is worth mentioning that the technical features of the sensor and other technical features involved in the present patent application should be regarded as prior art. The specific structure, working principle and possible control method and spatial arrangement method of these technical features can be selected conventionally in the art, and should not be regarded as the invention point of the present patent. The present patent will not be further expanded and detailed.

[0080] For those skilled in the art, the technical solutions described in the foregoing embodiments can still be modified, or some technical features can be replaced equivalently. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for optimizing sensor algorithm parameters based on non-convex optimization and entropy weighting, characterized in that, Includes the following steps: Step S1: Set the optimization objective function for the sensor and initialize the sensor parameters and data collection; Step S2: Optimize the objective function of the sensor using stochastic gradient descent. Step S3: Introduce the SiLU activation function to enhance the performance of the optimization process; Step S4: After completing the local parameter tuning, the entropy weight method is introduced to dynamically adjust the weights of the sensor nodes; Step S5: By combining centralized and distributed global data synchronization mechanisms, ensure that the optimization results are shared within a large-scale sensor network; For step S5: First, the central node collects information on latency and load from all nodes in the network to dynamically adjust the data synchronization frequency. It flexibly sets the synchronization interval through an adaptive control algorithm, thereby responding promptly to changes in data traffic and achieving efficient synchronization across the entire network based on node size and data traffic. In order to ensure the integrity of the synchronized data during data transmission, a packet loss detection and compensation mechanism is introduced. The central node monitors the transmission order and missing data packets. Once a packet loss is detected, a compensation request is triggered immediately, requiring the corresponding node to retransmit the lost data packets to ensure complete data transmission. In order to reduce the impact of packet loss and delay on system stability, a delay processing algorithm is adopted to dynamically adjust the delay of data packets. Delayed data is processed by priority sorting to ensure the timeliness of critical information. The synchronization mechanism combines centralized and distributed approaches: the central node is responsible for the overall network synchronization control and the aggregation of key data, while distributed nodes directly transmit local data to reduce dependence on the central node and improve synchronization response speed and stability. After each local parameter optimization is completed, each sensor node uploads the optimization results to the central node via wireless communication. The central node aggregates the optimization results of all nodes and calculates the globally optimal parameters. The central node feeds back the globally optimal parameters to each sensor node, ensuring that the parameters are optimized collaboratively on a global scale.

2. The sensor algorithm parameter tuning method based on non-convex optimization and entropy weighting method according to claim 1, characterized in that, For step S1: Assume the data collected by the sensor is The sensor parameters are The optimization objective is to minimize the non-convex objective function. The format is: ； in, This represents the weight of the i-th sensor. It is a non-convex function that represents the relationship between sensor performance and its parameters and data.

3. The sensor algorithm parameter tuning method based on non-convex optimization and entropy weighting method according to claim 2, characterized in that, For step S2: First, the parameters are randomly initialized, and then the gradient of the objective function is calculated. To update the sensor parameters, specifically: ； in, The learning rate controls the step size of parameter updates; after each parameter update, a projection operation is performed on the parameters to ensure that the parameters are within the feasible region, specifically: ； in, Let C be the projection function, and C be the feasible region of the constraints, which is related to the physical or application constraints of the sensor.

4. The sensor algorithm parameter tuning method based on non-convex optimization and entropy weighting method according to claim 3, characterized in that, For step S3: The formula for the SiLU activation function is: ； During gradient descent, the SiLU activation function is applied to the gradient update formula to optimize the nonlinear performance. The updated gradient formula is as follows: ； Gradient updates processed by this SiLU activation function effectively avoid getting trapped in local optima in non-convex optimization and improve the convergence speed of the global optimum.

5. The sensor algorithm parameter tuning method based on non-convex optimization and entropy weighting method according to claim 4, characterized in that, For step S4, the entropy weight method determines importance by calculating the rate of change of sensor data and dynamically adjusts the weight of each node based on the magnitude of information entropy. The specific calculation method is as follows: for each sensor node, the data change rate... Calculate information entropy : ； in, The information entropy represents the rate of change of data from the i-th sensor at time j. The higher the value, the greater the data change at that node, and the higher its priority is given in optimization; therefore, the node's weight is dynamically adjusted as follows: ； By using a dynamic adjustment mechanism to adaptively allocate weights in real time, nodes that experience drastic changes can respond to environmental changes more quickly, thereby improving the overall network tuning efficiency.

6. The sensor algorithm parameter tuning method based on non-convex optimization and entropy weight method according to claim 5, characterized in that, For step S5: Calculate the global optimal parameters The formula is: .

Images