WO2020051789A1 - 最小熵核密度估计器生成方法、装置和计算机可读存储介质 - Google Patents

最小熵核密度估计器生成方法、装置和计算机可读存储介质 Download PDF

Info

Publication number
WO2020051789A1
WO2020051789A1 PCT/CN2018/105174 CN2018105174W WO2020051789A1 WO 2020051789 A1 WO2020051789 A1 WO 2020051789A1 CN 2018105174 W CN2018105174 W CN 2018105174W WO 2020051789 A1 WO2020051789 A1 WO 2020051789A1
Authority
WO
WIPO (PCT)
Prior art keywords
window width
density estimator
kernel density
width parameter
value
Prior art date
Application number
PCT/CN2018/105174
Other languages
English (en)
French (fr)
Inventor
何玉林
蒋捷
Original Assignee
深圳大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳大学 filed Critical 深圳大学
Priority to PCT/CN2018/105174 priority Critical patent/WO2020051789A1/zh
Publication of WO2020051789A1 publication Critical patent/WO2020051789A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10Complex mathematical operations
    • G06F17/18Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis

Definitions

  • the present application belongs to the field of data mining, and particularly relates to a method, a device, and a computer-readable storage medium for generating a minimum entropy kernel density estimator.
  • the classical probability density function estimation method is the kernel density estimation method, also known as the Parzen window method. It uses the average of the kernel function superposition to fit the true probability density function of the data to generate a kernel density estimator to estimate the unknown distribution data. Probability density function.
  • the key lies in the selection of the window width parameter: a larger window width will result in "oversmoothing" density estimation, and a smaller window width will result in "undersmoothing" density estimation.
  • an objective function based on minimizing the true density and estimated density error is generally required.
  • BCV-KDE BiasedCross-ValidationBasedKernelDensityEstimator
  • BCV-KDE builds an objective function that calculates the optimal window width parameters of the kernel density estimator based on a cross-validation strategy, and then determines the optimal window width parameters of the kernel density estimator through optimization algorithms such as gradient descent, fixed point iteration, and particle swarm optimization .
  • the purpose of this application is to provide a method, a device and a computer-readable storage medium for generating a minimum entropy kernel density estimator, so as to reduce the uncertainty of the objective function and enhance the robustness of the kernel density estimator.
  • a first aspect of the present application provides a method for generating a minimum entropy kernel density estimator, where the method includes:
  • the objective function uses a second kernel density estimator as a variable of a logarithmic function, and the second kernel density estimator is constructed for the second data set,
  • the second data set is formed by excluding any one data from the first data set;
  • a second aspect of the present application provides a device for generating a minimum entropy kernel density estimator, where the device includes:
  • a construction module configured to construct a first kernel density estimator for the first data set, where the first kernel density estimator includes a window width parameter h having a value greater than 0;
  • a determining module configured to determine an objective function for obtaining an optimal value of the window width parameter h, where the objective function uses a second kernel density estimator as a variable of a logarithmic function, and the second kernel density estimator Construction of two data sets, the second data set is formed by excluding any one data from the first data set;
  • Iterative module for selecting a preset error threshold ⁇ and an initial value h 0 of the window width parameter h, and using an iterative algorithm to iterate the value of the window width parameter h to an absolute difference from the value of h obtained in the previous iteration It is not more than ⁇ .
  • a third aspect of the present application provides a computing device including a memory, a processor, and a computer program stored in the memory and executable on the processor.
  • the processor executes the computer program, the following method is implemented: step:
  • the objective function uses a second kernel density estimator as a variable of a logarithmic function, and the second kernel density estimator is constructed for the second data set,
  • the second data set is formed by excluding any one data from the first data set;
  • a fourth aspect of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the following method are implemented:
  • the objective function uses a second kernel density estimator as a variable of a logarithmic function, and the second kernel density estimator is constructed for the second data set,
  • the second data set is formed by excluding any one data from the first data set;
  • the second kernel density estimator is The construction of the second data set, that is, the objective function depends on a given data set.
  • the technical solution of the present application reduces The uncertainty of the objective function used to obtain the optimal value of the window width parameter h; on the other hand, a preset error threshold ⁇ and the initial value h 0 of the window width parameter h are selected, and the window width parameter h is The value is iterated until the absolute difference between the value of h obtained in the previous iteration is not greater than ⁇ , which makes the time overhead of determining the optimal value of the window width parameter h smaller, and also helps to enhance the robustness of the kernel density estimator.
  • FIG. 1 is a schematic flowchart of an implementation method for generating a minimum entropy kernel density estimator according to an embodiment of the present application
  • FIG. 2 is a related schematic diagram of a minimum entropy kernel density estimator generated by a minimum entropy kernel density estimator generation method provided in an embodiment of the present application;
  • FIG. 3 is a related schematic diagram of a minimum entropy kernel density estimator generated by a method for generating a minimum entropy kernel density estimator according to another embodiment of the present application;
  • FIG. 4 is a schematic structural diagram of a minimum entropy kernel density estimator generating device according to an embodiment of the present application
  • FIG. 5 is a schematic structural diagram of a minimum entropy kernel density estimator generating apparatus according to another embodiment of the present application.
  • FIG. 6 is a schematic structural diagram of a computing device according to an embodiment of the present application.
  • FIG. 1 is a schematic flowchart of an implementation method for generating a minimum entropy kernel density estimator provided by an embodiment of the present application, which mainly includes the following steps S101 to S104, which are described in detail below:
  • D ⁇ x i
  • x i ⁇ R, i 1,2, ..., L, ..., N ⁇
  • D ⁇ x i
  • x i ⁇ R, i 1,2, ..., L, ..., N ⁇
  • R represents a real number set
  • N represents the number of data x i in the first data set
  • h represents a window width parameter, Gaussian kernel function
  • the objective function uses a second kernel density estimator as a logarithmic function.
  • the second kernel density estimator is constructed for the second data set.
  • the data set is formed by excluding any data from the first data set.
  • the key of the probability density function estimation is to select an optimal value of the window width parameter h, that is, the optimal window width parameter.
  • the objective function of the optimal value of the window width parameter h can be designed as:
  • a second kernel density estimator constructed for the second data set D -i Dx i , Is the probability density value corresponding to the data x i , and the second data set here is formed by excluding any one data from the first data set.
  • step S103 the Substitute the objective function to solve Get the following expression for the window width parameter h:
  • step S104 can be implemented by the following steps S'1 to S'5:
  • the objective function used to obtain the optimal value of the window width parameter h uses the second kernel density estimator as a logarithmic function .
  • the second kernel density estimator is constructed for the second data set, that is, the objective function depends on the given data set, so it is inconsistent with the increase in the uncertainty of the objective function caused by the introduction of unknown quantities into the unknown in the prior art
  • the technical solution of the present application reduces the uncertainty of the objective function used to obtain the optimal value of the window width parameter h; on the other hand, a preset error threshold ⁇ and an initial value h 0 of the window width parameter h are selected.
  • the iterative algorithm is used to iterate the value of the window width parameter h to an absolute difference between the value of h obtained in the previous iteration and not more than ⁇ , so that the time overhead for determining the optimal value of the window width parameter h is smaller, and it is also conducive to enhancement
  • the circles in the middle graph represent 200 random numbers that follow the standard normal distribution.
  • the left graph is Corresponding curve, where the window width parameter h corresponding to the minimum entropy is 0.473; the right figure is a learning curve for optimizing the window width parameter h using the technical scheme of FIG. 1.
  • the circles in the middle graph represent 200 random numbers that obey the standard normal distribution.
  • the left graph is Corresponding curve, where the window width parameter h corresponding to the minimum entropy is 0.350; the right figure is a learning curve for optimizing the window width parameter h using the technical scheme of FIG. 1.
  • FIG. 4 is a schematic diagram of a minimum entropy kernel density estimator generating apparatus provided by an embodiment of the present application. For ease of explanation, only necessary parts of the device are shown.
  • the device illustrated in FIG. 4 mainly includes a construction module 401, a determination module 402, a solution module 403, and an iteration module 404, which are described in detail as follows:
  • a construction module 401 is configured to construct a first kernel density estimator for the first data set, where the first kernel density estimator includes a window width parameter h having a value greater than 0;
  • a determining module 402 configured to determine an objective function for obtaining an optimal value of the window width parameter h, wherein the objective function uses a second kernel density estimator as a logarithmic function variable, and the second kernel density estimator is directed to the second data set Constructing, the second data set is formed by excluding any one data from the first data set;
  • a solving module 403, configured to solve the expression of the window width parameter h by the second kernel density estimator and the objective function
  • Iteration module 404 for selecting a preset error threshold ⁇ and an initial value h 0 of the window width parameter h, and using an iterative algorithm to iterate the value of the window width parameter h to an absolute difference between the value of h obtained in the previous iteration and not greater than ⁇ until.
  • the construction module 401 illustrated in FIG. 4 may include an estimator construction unit 501, such as a minimum entropy kernel density estimator generating device illustrated in FIG. 5.
  • x i ⁇ R, i 1,2, ..., L, ..., N ⁇ as the first data set to construct a kernel density estimation in the following form Device
  • R represents a real number set
  • N represents the number of data of the first data set
  • h represents a window width parameter
  • the objective function of the optimal value of the window width parameter h is:
  • a second kernel density estimator constructed for the second data set D -i Dx i , Is the probability density value corresponding to the data x i .
  • FIG. 6 is a schematic structural diagram of a computing device according to an embodiment of the present application.
  • the computing device 6 of this embodiment includes a processor 60, a memory 61, and a computer program 62 stored in the memory 61 and executable on the processor 60, such as a method for generating a minimum entropy kernel density estimator. program.
  • the processor 60 executes the computer program 62
  • the steps in the embodiment of the method for generating the minimum entropy kernel density estimator described above are implemented, for example, steps S101 to S104 shown in FIG. 1, or when the processor 60 executes the computer program 62, the foregoing devices are implemented.
  • the functions of each module / unit in the example are, for example, the functions of the building module 401, the determination module 402, the solving module 403, and the iteration module 404 shown in FIG.
  • the computer program 62 of the method for generating the minimum entropy kernel density estimator mainly includes: constructing a first kernel density estimator for a first data set, the first kernel density estimator containing a window width parameter h greater than 0; Determine an objective function for obtaining an optimal value of the window width parameter h, the objective function uses a second kernel density estimator as a variable of a logarithmic function, and the second kernel density estimator is constructed for the second data set, The second data set is formed by excluding any data from the first data set; the expression of the window width parameter h is solved by the second kernel density estimator and an objective function; a preset error threshold ⁇ and An initial value h 0 of the window width parameter h is iterated using an iterative algorithm to an absolute difference between the value of the window width parameter h and the value of h obtained in the previous iteration is not greater than the ⁇ .
  • the computer program 62 may be divided into one or more modules / units, and one or more modules / units are stored in the memory 61 and executed by the processor 60 to complete the present application.
  • One or more modules / units may be a series of computer program instruction segments capable of performing a specific function, and the instruction segments are used to describe the execution process of the computer program 62 in the computing device 6.
  • the computer program 62 may be divided into functions of a building module 401, a determining module 402, a solving module 403, and an iterative module 404 (modules in a virtual device).
  • each module is used for the first data Set to build a first kernel density estimator, where the first kernel density estimator contains a window width parameter h with a value greater than 0; a determination module 402 is used to determine an objective function for obtaining an optimal value of the window width parameter h, where: The objective function uses a second kernel density estimator as a logarithmic function.
  • the second kernel density estimator is constructed for a second data set, and the second data set is formed by excluding any data from the first data set; a solving module 403 , Used to solve the expression of the window width parameter h by the second kernel density estimator and the objective function; the iteration module 404 is used to select a preset error threshold ⁇ and the initial value h 0 of the window width parameter h, and use an iterative algorithm to convert the window The value of the width parameter h is iterated until the absolute difference from the value of h obtained in the previous iteration is not greater than ⁇ .
  • the computing device 6 may include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art can understand that FIG. 6 is only an example of the computing device 6 and does not constitute a limitation on the computing device 6. It may include more or fewer components than shown in the figure, or combine some components, or different components. For example, computing devices may also include input and output devices, network access devices, and buses.
  • the so-called processor 60 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), Ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • a general-purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
  • the memory 61 may be an internal storage unit of the computing device 6, such as a hard disk or a memory of the computing device 6.
  • the memory 61 may also be an external storage device of the computing device 6, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, and a flash memory card (Flash) provided on the computing device 6. Card) and so on.
  • the memory 61 may also include both an internal storage unit of the computing device 6 and an external storage device.
  • the memory 61 is used to store computer programs and other programs and data required by the computing device.
  • the memory 61 may also be used to temporarily store data that has been output or is to be output.
  • the disclosed apparatus / computing device and method may be implemented in other ways.
  • the device / computing device embodiments described above are only schematic.
  • the division of modules or units is only a logical function division.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, which may be electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, which may be located in one place, or may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objective of the solution of this embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, or each of the units may exist separately physically, or two or more units may be integrated into one unit.
  • the above integrated unit may be implemented in the form of hardware or in the form of software functional unit.
  • Integrated modules / units can be stored in a computer-readable storage medium if they are implemented in the form of software functional units and sold or used as independent products. Based on this understanding, this application implements all or part of the processes in the methods of the above embodiments, and can also be completed by computer programs instructing related hardware.
  • the computer program of the method for generating the minimum entropy kernel density estimator can be stored in a computer In a storage medium, when the computer program is executed by a processor, the steps of the foregoing method embodiments may be implemented, that is, a first kernel density estimator is constructed for a first data set, and the first kernel density estimator contains a value greater than A window width parameter h of 0; determining an objective function for obtaining an optimal value of the window width parameter h, the objective function using a second kernel density estimator as a logarithmic function variable, and the second kernel density estimator For the second data set, the second data set is formed by excluding any one data from the first data set; the expression of the window width parameter h is solved by the second kernel density estimator and an objective function; selecting a preset error threshold ⁇ and the window width parameter h initial value h 0, h values of the window width parameter using an iterative algorithm to the previous iteration H iterations resulting absolute difference value is not larger than the ⁇ far.
  • the computer program includes computer program code, and the computer program code may be in a source code form, an object code form, an executable file, or some intermediate form.
  • the computer-readable medium may include: any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM, Read-Only Memory), random access Memory (RAM, Random Access Memory), electric carrier signals, telecommunication signals, and software distribution media. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of the legislation and patent practice in the jurisdiction.
  • the computer-readable medium does not include Electric carrier signals and telecommunication signals.
  • the above embodiments are only used to describe the technical solution of the present application, but not limited thereto.
  • the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still apply the foregoing embodiments.
  • the recorded technical solutions are modified, or some of the technical features are equivalently replaced; and these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the application, and should be included in the application.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Operations Research (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Algebra (AREA)
  • Evolutionary Biology (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Complex Calculations (AREA)

Abstract

一种最小熵核密度估计器生成方法、装置和计算机可读存储介质,以减小目标函数的不确定性和增强核密度估计器的健壮性。所述方法包括:针对第一数据集构建第一核密度估计器;确定求取所述窗口宽度参数h的最优值的目标函数;由第二核密度估计器和目标函数求解窗口宽度参数h的表达式;选取预设误差阈值ξ和窗口宽度参数h的初始值h 0,采用迭代算法将窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于ξ为止。本申请的技术方案减小了目标函数的不确定性和增强了核密度估计器的健壮性。

Description

最小熵核密度估计器生成方法、装置和计算机可读存储介质 技术领域
本申请属于数据挖掘领域,尤其涉及一种最小熵核密度估计器生成方法、装置和计算机可读存储介质。
背景技术
估计未知分布数据的概率密度函数是数据挖掘领域的一项重要研究内容。经典的概率密度函数估计方法是核密度估计方法,又称Parzen窗口法,它利用核函数叠加的平均去拟合数据真实的概率密度函数,从而生成核密度估计器,以此估计未知分布数据的概率密度函数。
对于核密度估计器的构建而言,关键在于窗口宽度参数的选取:较大的窗口宽度将导致“过平滑”的密度估计,而较小的窗口宽度将导致“欠平滑”的密度估计。对于最优窗口宽度参数的选取,一般需要构建一个基于真实密度和估计密度误差最小化的目标函数。其中,有偏交叉验证核密度估计器(BiasedCross-ValidationBasedKernel Density Estimator,BCV-KDE)是一种经典的未知分布数据集概率密度函数的估计器。BCV-KDE基于交叉验证的策略构建计算核密度估计器最优窗口宽度参数的目标函数,之后通过诸如梯度下降、不动点迭代、粒子群等优化算法确定核密度估计器的最优窗口宽度参数。
在BCV-KDE的目标函数中,真实密度是未知的,因此不得不通过交叉验证的策略去近似数据集的真实密度。而目标函数中已存在一个估计量(即估计密度),为了近似真实密度又必须引入新的估计量,这就相当于在未知中再次引入未知,从而增加了目标函数的不确定性和降低了核密度估计器的健壮性。
发明内容
本申请的目的在于提供一种最小熵核密度估计器生成方法、装置和计算机 可读存储介质,以减小目标函数的不确定性和增强核密度估计器的健壮性。
本申请第一方面提供一种最小熵核密度估计器生成方法,所述方法包括:
针对第一数据集构建第一核密度估计器,所述第一核密度估计器含有值大于0的窗口宽度参数h;
确定求取所述窗口宽度参数h的最优值的目标函数,所述目标函数以第二核密度估计器为对数函数的变量,所述第二核密度估计器针对第二数据集构建,所述第二数据集为从所述第一数据集剔除任意一个数据构成;
由所述第二核密度估计器和目标函数求解所述窗口宽度参数h的表达式;
选取预设误差阈值ξ和所述窗口宽度参数h的初始值h 0,采用迭代算法将所述窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于所述ξ为止。
本申请第二方面提供一种最小熵核密度估计器生成装置,所述装置包括:
构建模块,用于针对第一数据集构建第一核密度估计器,所述第一核密度估计器含有值大于0的窗口宽度参数h;
确定模块,用于确定求取所述窗口宽度参数h的最优值的目标函数,所述目标函数以第二核密度估计器为对数函数的变量,所述第二核密度估计器针对第二数据集构建,所述第二数据集为从所述第一数据集剔除任意一个数据构成;
求解模块,用于由所述第二核密度估计器和目标函数求解所述窗口宽度参数h的表达式;
迭代模块,用于选取预设误差阈值ξ和所述窗口宽度参数h的初始值h 0,采用迭代算法将所述窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于所述ξ为止。
本申请第三方面提供一种计算设备,包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序,所述处理器执行所述计算机程序时实现如下方法的步骤:
针对第一数据集构建第一核密度估计器,所述第一核密度估计器含有值大于0的窗口宽度参数h;
确定求取所述窗口宽度参数h的最优值的目标函数,所述目标函数以第二核密度估计器为对数函数的变量,所述第二核密度估计器针对第二数据集构建, 所述第二数据集为从所述第一数据集剔除任意一个数据构成;
由所述第二核密度估计器和目标函数求解所述窗口宽度参数h的表达式;
选取预设误差阈值ξ和所述窗口宽度参数h的初始值h 0,采用迭代算法将所述窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于所述ξ为止。
本申请第四方面提供一种计算机可读存储介质,所述计算机可读存储介质存储有计算机程序,所述计算机程序被处理器执行时实现如下方法的步骤:
针对第一数据集构建第一核密度估计器,所述第一核密度估计器含有值大于0的窗口宽度参数h;
确定求取所述窗口宽度参数h的最优值的目标函数,所述目标函数以第二核密度估计器为对数函数的变量,所述第二核密度估计器针对第二数据集构建,所述第二数据集为从所述第一数据集剔除任意一个数据构成;
由所述第二核密度估计器和目标函数求解所述窗口宽度参数h的表达式;
选取预设误差阈值ξ和所述窗口宽度参数h的初始值h 0,采用迭代算法将所述窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于所述ξ为止。
从上述本申请技术方案可知,一方面,由于,用于求取窗口宽度参数h的最优值的目标函数以第二核密度估计器为对数函数的变量,而第二核密度估计器针对第二数据集构建,即目标函数依赖于给定的数据集,因此,与现有技术在未知量中再次引入未知量导致的目标函数的不确定性增加相比,本申请的技术方案降低了用于求取窗口宽度参数h的最优值的目标函数的不确定性;另一方面,选取预设误差阈值ξ和窗口宽度参数h的初始值h 0,采用迭代算法将窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于ξ为止,使得确定窗口宽度参数h的最优值的时间开销更小,也使得有利于增强核密度估计器的健壮性。
附图说明
图1是本申请实施例提供的最小熵核密度估计器生成方法的实现流程示意图;
图2本申请实施例提供的最小熵核密度估计器生成方法生成的最小熵核密度估计器的相关示意图;
图3本申请另一实施例提供的最小熵核密度估计器生成方法生成的最小熵核密度估计器的相关示意图;
图4本申请实施例提供的最小熵核密度估计器生成装置的结构示意图;
图5是本申请另一实施例提供的最小熵核密度估计器生成装置的结构示意图;
图6是本申请实施例提供的计算设备的结构示意图。
具体实施方式
为了使本申请的目的、技术方案及有益效果更加清楚明白,以下结合附图及实施例,对本申请进行进一步详细说明。应当理解,此处所描述的具体实施例仅仅用以解释本申请,并不用于限定本申请。
以下描述中,为了说明而不是为了限定,提出了诸如特定系统结构、技术之类的具体细节,以便透彻理解本申请实施例。然而,本领域的技术人员应当清楚,在没有这些具体细节的其它实施例中也可以实现本申请。在其它情况中,省略对众所周知的系统、装置、电路以及方法的详细说明,以免不必要的细节妨碍本申请的描述。
附图1是本申请实施例提供的最小熵核密度估计器生成方法的实现流程示意图,主要包括以下步骤S101至S104,以下详细说明:
S101,针对第一数据集构建第一核密度估计器,其中,第一核密度估计器含有值大于0的窗口宽度参数h。
具体地,可以以D={x i|x i∈R,i=1,2,...,L,...,N}为第一数据集,构建如下形式的核密度估计器
Figure PCTCN2018105174-appb-000001
Figure PCTCN2018105174-appb-000002
其中,R表示实数集,N表示第一数据集中的数据x i的个数,h表示窗口宽度参数,
Figure PCTCN2018105174-appb-000003
是高斯核函数
Figure PCTCN2018105174-appb-000004
S102,确定求取窗口宽度参数h的最优值的目标函数,其中,目标函数以第二核密度估计器为对数函数的变量,第二核密度估计器针对第二数据集构建,第二数据集为从第一数据集剔除任意一个数据构成。
本领域技术人员可以理解的是,概率密度函数估计即核密度估计器的关键在于选取一个窗口宽度参数h的最优值即最优窗口宽度参数。为了选取最优窗口宽度参数,窗口宽度参数h的最优值的目标函数可以设计为:
Figure PCTCN2018105174-appb-000005
其中,
Figure PCTCN2018105174-appb-000006
为针对第二数据集D -i=D-x i构建的第二核密度估计器,
Figure PCTCN2018105174-appb-000007
为数据x i对应的概率密度值,此处的第二数据集是从第一数据集剔除任意一个数据构成。
S103,由第二核密度估计器和窗口宽度参数h的最优值的目标函数求解窗口宽度参数h的表达式。
为了实现步骤S103,可以将
Figure PCTCN2018105174-appb-000008
代入目标函数,进而求解
Figure PCTCN2018105174-appb-000009
得到窗口宽度参数h的如下表达式:
Figure PCTCN2018105174-appb-000010
其中,
Figure PCTCN2018105174-appb-000011
具体地,将
Figure PCTCN2018105174-appb-000012
代入目标函数
Figure PCTCN2018105174-appb-000013
时,可以得到:
Figure PCTCN2018105174-appb-000014
进而对h求导并令
Figure PCTCN2018105174-appb-000015
可以得到:
Figure PCTCN2018105174-appb-000016
即:
Figure PCTCN2018105174-appb-000017
S104,选取预设误差阈值ξ和窗口宽度参数h的初始值h 0,采用迭代算法将窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于ξ为止。
由于从上述表达式(6)很难求解出窗口宽度参数h的最优值的显示解,因此,步骤S104可通过如下步骤S’1至S’5实现:
S’1,将初始值h0代入
Figure PCTCN2018105174-appb-000018
中等号右边的
Figure PCTCN2018105174-appb-000019
求得h的一个值h k
S’2,比较|h k-h 0|与ξ的大小;
S’3,若|h k-h 0|不大于ξ,则将h 0作为窗口宽度参数h的最优值并停止迭代 算法,若|h k-h 0|大于ξ,则将h k代入
Figure PCTCN2018105174-appb-000020
中等号右边的
Figure PCTCN2018105174-appb-000021
求得h的一个值h' k
S’4,比较|h' k-h 0|与ξ的大小;
S’5,若|h' k-h 0|不大于ξ,则将h' k作为窗口宽度参数h的最优值并停止迭代算法,若|h' k-h 0|大于ξ,则将h' k代入
Figure PCTCN2018105174-appb-000022
中等号右边的
Figure PCTCN2018105174-appb-000023
将求得h的值记为h' k
重复上述步骤S’4和S’5,直到|h' k-h 0|不大于ξ,将h' k作为窗口宽度参数h的最优值并停止迭代算法。
从上述附图1示例的最小熵核密度估计器生成方法可知,一方面,由于,用于求取窗口宽度参数h的最优值的目标函数以第二核密度估计器为对数函数的变量,而第二核密度估计器针对第二数据集构建,即目标函数依赖于给定的数据集,因此,与现有技术在未知量中再次引入未知量导致的目标函数的不确定性增加相比,本申请的技术方案降低了用于求取窗口宽度参数h的最优值的目标函数的不确定性;另一方面,选取预设误差阈值ξ和窗口宽度参数h的初始值h 0,采用迭代算法将窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于ξ为止,使得确定窗口宽度参数h的最优值的时间开销更小,也使得有利于增强核密度估计器的健壮性。
为了更好地说明上述本申请技术方案的有益效果,以下举例说明之:
如附图2所示,中间图的圆圈表示200个服从标准正态分布的随机数,左图为
Figure PCTCN2018105174-appb-000024
对应的曲线,其中最小熵对应的窗口宽度参数h为0.473;右图为使用附 图1的技术方案优化窗口宽度参数h的学习曲线,经过93次迭代,窗口宽度参数h从初始值h 0=5收敛到最优值即最优窗口宽度参数
Figure PCTCN2018105174-appb-000025
其中,预设误差阈值ξ=10 -5;中图黑线即为采用了
Figure PCTCN2018105174-appb-000026
的ME-KDE。
再如附图3所示,中间图的圆圈表示200个服从标准正态分布的随机数,左图为
Figure PCTCN2018105174-appb-000027
对应的曲线,其中最小熵对应的窗口宽度参数h为0.350;右图为使用附图1的技术方案优化窗口宽度参数h的学习曲线,经过180次迭代,窗口宽度参数h从初始值h 0=0.001收敛到最优值即最优窗口宽度参数
Figure PCTCN2018105174-appb-000028
其中,预设误差阈值ξ=10 -5;中图黑线即为采用了
Figure PCTCN2018105174-appb-000029
的ME-KDE。
图4是本申请实施例提供的最小熵核密度估计器生成装置的示意图。为了便于说明,仅仅示出了装置的必要部分。图4示例的装置主要包括构建模块401、确定模块402、求解模块403和迭代模块404,详细说明如下:
构建模块401,用于针对第一数据集构建第一核密度估计器,其中,第一核密度估计器含有值大于0的窗口宽度参数h;
确定模块402,用于确定求取窗口宽度参数h的最优值的目标函数,其中,目标函数以第二核密度估计器为对数函数的变量,第二核密度估计器针对第二数据集构建,第二数据集为从所述第一数据集剔除任意一个数据构成;
求解模块403,用于由第二核密度估计器和目标函数求解窗口宽度参数h的表达式;
迭代模块404,用于选取预设误差阈值ξ和窗口宽度参数h的初始值h 0,采用迭代算法将窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于ξ为止。
需要说明的是,本申请实施例提供的装置,由于与本申请方法实施例基于同一构思,其带来的技术效果与本申请方法实施例相同,具体内容可参见本申请方法实施例中的叙述,此处不再赘述。
附图4示例的构建模块401可以包括估计器构建单元501,如附图5示例的最小熵核密度估计器生成装置。估计器构建单元501用于以D={x i|x i∈R,i=1,2,...,L,...,N}为第一数据集,构建如下形式的核密度估计器
Figure PCTCN2018105174-appb-000030
Figure PCTCN2018105174-appb-000031
其中,R表示实数集,N表示第一数据集的 数据个数,h表示窗口宽度参数,
Figure PCTCN2018105174-appb-000032
是高斯核函数
Figure PCTCN2018105174-appb-000033
上述附图4或5示例的最小熵核密度估计器生成装置中,窗口宽度参数h的最优值的目标函数为:
Figure PCTCN2018105174-appb-000034
为针对第二数据集D -i=D-x i构建的第二核密度估计器,
Figure PCTCN2018105174-appb-000035
为数据x i对应的概率密度值。
图6是本申请一实施例提供的计算设备的结构示意图。如图6所示,该实施例的计算设备6包括:处理器60、存储器61以及存储在存储器61中并可在处理器60上运行的计算机程序62,例如最小熵核密度估计器生成方法的程序。处理器60执行计算机程序62时实现上述最小熵核密度估计器生成方法实施例中的步骤,例如图1所示的步骤S101至S104,或者,处理器60执行计算机程序62时实现上述各装置实施例中各模块/单元的功能,例如图4所示构建模块401、确定模块402、求解模块403和迭代模块404的功能。
示例性的,最小熵核密度估计器生成方法的计算机程序62主要包括:针对第一数据集构建第一核密度估计器,所述第一核密度估计器含有值大于0的窗口宽度参数h;确定求取所述窗口宽度参数h的最优值的目标函数,所述目标函数以第二核密度估计器为对数函数的变量,所述第二核密度估计器针对第二数据集构建,所述第二数据集为从所述第一数据集剔除任意一个数据构成;由所述第二核密度估计器和目标函数求解所述窗口宽度参数h的表达式;选取预设误差阈值ξ和所述窗口宽度参数h的初始值h 0,采用迭代算法将所述窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于所述ξ为止。计算机程序62可以被分割成一个或多个模块/单元,一个或者多个模块/单元被存储在存储器61中,并由处理器60执行,以完成本申请。一个或多个模块/单元可以是能够完成特定功能的一系列计算机程序指令段,该指令段用于描述计算机程序62在计算设备6中的执行过程。例如,计算机程序62可以被分割成构建模块401、确定模块402、求解模块403和迭代模块404(虚拟装置中的 模块)的功能,各模块具体功能如下:构建模块401,用于针对第一数据集构建第一核密度估计器,其中,第一核密度估计器含有值大于0的窗口宽度参数h;确定模块402,用于确定求取窗口宽度参数h的最优值的目标函数,其中,目标函数以第二核密度估计器为对数函数的变量,第二核密度估计器针对第二数据集构建,第二数据集为从所述第一数据集剔除任意一个数据构成;求解模块403,用于由第二核密度估计器和目标函数求解窗口宽度参数h的表达式;迭代模块404,用于选取预设误差阈值ξ和窗口宽度参数h的初始值h 0,采用迭代算法将窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于ξ为止。
计算设备6可包括但不仅限于处理器60、存储器61。本领域技术人员可以理解,图6仅仅是计算设备6的示例,并不构成对计算设备6的限定,可以包括比图示更多或更少的部件,或者组合某些部件,或者不同的部件,例如计算设备还可以包括输入输出设备、网络接入设备、总线等。
所称处理器60可以是中央处理单元(Central Processing Unit,CPU),还可以是其他通用处理器、数字信号处理器(Digital Signal Processor,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现成可编程门阵列(Field-Programmable Gate Array,FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
存储器61可以是计算设备6的内部存储单元,例如计算设备6的硬盘或内存。存储器61也可以是计算设备6的外部存储设备,例如计算设备6上配备的插接式硬盘,智能存储卡(Smart Media Card,SMC),安全数字(Secure Digital,SD)卡,闪存卡(Flash Card)等。进一步地,存储器61还可以既包括计算设备6的内部存储单元也包括外部存储设备。存储器61用于存储计算机程序以及计算设备所需的其他程序和数据。存储器61还可以用于暂时地存储已经输出或者将要输出的数据。
所属领域的技术人员可以清楚地了解到,为了描述的方便和简洁,仅以上述各功能单元、模块的划分进行举例说明,实际应用中,可以根据需要而将上述功能分配由不同的功能单元、模块完成,即将装置的内部结构划分成不同的 功能单元或模块,以完成以上描述的全部或者部分功能。实施例中的各功能单元、模块可以集成在一个处理单元中,也可以是各个单元单独物理存在,也可以两个或两个以上单元集成在一个单元中,上述集成的单元既可以采用硬件的形式实现,也可以采用软件功能单元的形式实现。另外,各功能单元、模块的具体名称也只是为了便于相互区分,并不用于限制本申请的保护范围。上述系统中单元、模块的具体工作过程,可以参考前述方法实施例中的对应过程,在此不再赘述。
在上述实施例中,对各个实施例的描述都各有侧重,某个实施例中没有详述或记载的部分,可以参见其它实施例的相关描述。
本领域普通技术人员可以意识到,结合本文中所公开的实施例描述的各示例的单元及算法步骤,能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行,取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能,但是这种实现不应认为超出本申请的范围。
在本申请所提供的实施例中,应该理解到,所揭露的装置/计算设备和方法,可以通过其它的方式实现。例如,以上所描述的装置/计算设备实施例仅仅是示意性的,例如,模块或单元的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式,例如多个单元或组件可以结合或者可以集成到另一个系统,或一些特征可以忽略,或不执行。另一点,所显示或讨论的相互之间的耦合或直接耦合或通讯连接可以是通过一些接口,装置或单元的间接耦合或通讯连接,可以是电性,机械或其它的形式。
作为分离部件说明的单元可以是或者也可以不是物理上分开的,作为单元显示的部件可以是或者也可以不是物理单元,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外,在本申请各个实施例中的各功能单元可以集成在一个处理单元中,也可以是各个单元单独物理存在,也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现,也可以采用软件功能单元的形式实现。
集成的模块/单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时,可以存储在一个计算机可读取存储介质中。基于这样的理解,本申请实现上述实施例方法中的全部或部分流程,也可以通过计算机程序来指令相关的硬件来完成,最小熵核密度估计器生成方法的计算机程序可存储于一计算机可读存储介质中,该计算机程序在被处理器执行时,可实现上述各个方法实施例的步骤,即,针对第一数据集构建第一核密度估计器,所述第一核密度估计器含有值大于0的窗口宽度参数h;确定求取所述窗口宽度参数h的最优值的目标函数,所述目标函数以第二核密度估计器为对数函数的变量,所述第二核密度估计器针对第二数据集构建,所述第二数据集为从所述第一数据集剔除任意一个数据构成;由所述第二核密度估计器和目标函数求解所述窗口宽度参数h的表达式;选取预设误差阈值ξ和所述窗口宽度参数h的初始值h 0,采用迭代算法将所述窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于所述ξ为止。其中,计算机程序包括计算机程序代码,计算机程序代码可以为源代码形式、对象代码形式、可执行文件或某些中间形式等。计算机可读介质可以包括:能够携带计算机程序代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM,Read-Only Memory)、随机存取存储器(RAM,Random Access Memory)、电载波信号、电信信号以及软件分发介质等。需要说明的是,计算机可读介质包含的内容可以根据司法管辖区内立法和专利实践的要求进行适当的增减,例如在某些司法管辖区,根据立法和专利实践,计算机可读介质不包括电载波信号和电信信号。以上实施例仅用以说明本申请的技术方案,而非对其限制;尽管参照前述实施例对本申请进行了详细的说明,本领域的普通技术人员应当理解:其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围,均应包含在本申请的保护范围之内。

Claims (10)

  1. 一种最小熵核密度估计器生成方法,其特征在于,所述方法包括:
    针对第一数据集构建第一核密度估计器,所述第一核密度估计器含有值大于0的窗口宽度参数h;
    确定求取所述窗口宽度参数h的最优值的目标函数,所述目标函数以第二核密度估计器为对数函数的变量,所述第二核密度估计器针对第二数据集构建,所述第二数据集为从所述第一数据集剔除任意一个数据构成;
    由所述第二核密度估计器和目标函数求解所述窗口宽度参数h的表达式;
    选取预设误差阈值ξ和所述窗口宽度参数h的初始值h 0,采用迭代算法将所述窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于所述ξ为止。
  2. 如权利要求1所述的最小熵核密度估计器生成方法,其特征在于,所述针对第一数据集构建第一核密度估计器,包括:
    以D={x i|x i∈R,i=1,2,...,L,...,N}为所述第一数据集,构建如下形式的核密度估计器
    Figure PCTCN2018105174-appb-100001
    Figure PCTCN2018105174-appb-100002
    所述R表示实数集,所述N表示所述第一数据集的数据个数,所述h表示所述窗口宽度参数,所述
    Figure PCTCN2018105174-appb-100003
    是高斯核函数
    Figure PCTCN2018105174-appb-100004
  3. 如权利要求1或2所述的最小熵核密度估计器生成方法,其特征在于,所述窗口宽度参数h的最优值的目标函数为:
    Figure PCTCN2018105174-appb-100005
    其中,
    Figure PCTCN2018105174-appb-100006
    为针对所述第二数据集D -i=D-x i构建的第二核密度估计器,
    Figure PCTCN2018105174-appb-100007
    为数据x i对应的概率密度值。
  4. 如权利要求3所述的最小熵核密度估计器生成方法,其特征在于,所述 由所述第二核密度估计器和目标函数求解所述窗口宽度参数h的表达式,包括:
    将所述
    Figure PCTCN2018105174-appb-100008
    代入所述目标函数,进而求解
    Figure PCTCN2018105174-appb-100009
    得到所述窗口宽度参数h的如下表达式:
    Figure PCTCN2018105174-appb-100010
    其中,
    Figure PCTCN2018105174-appb-100011
  5. 如权利要求4所述的最小熵核密度估计器生成方法,其特征在于,所述选取预设误差阈值ξ和所述窗口宽度参数h的初始值h 0,采用迭代算法将所述窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于所述ξ为止,包括:
    S’1,将初始值h 0代入
    Figure PCTCN2018105174-appb-100012
    中等号右边的
    Figure PCTCN2018105174-appb-100013
    求得所述h的一个值h k
    S’2,比较|h k-h 0|与ξ的大小;
    S’3,若|h k-h 0|不大于ξ,则将h 0作为所述窗口宽度参数h的最优值并停止所述迭代算法,若|h k-h 0|大于ξ,则将h k代入
    Figure PCTCN2018105174-appb-100014
    中等号右边的
    Figure PCTCN2018105174-appb-100015
    求得所述h的一个值h' k
    S’4,比较|h' k-h 0|与ξ的大小;
    S’5,若|h' k-h 0|不大于ξ,则将h' k作为所述窗口宽度参数h的最优值并停止 所述迭代算法,若|h' k-h 0|大于ξ,则将h' k代入
    Figure PCTCN2018105174-appb-100016
    中等号右边的
    Figure PCTCN2018105174-appb-100017
    将求得h的值记为h' k
    重复上述步骤S’4和S’5。
  6. 一种最小熵核密度估计器生成装置,其特征在于,所述装置包括:
    构建模块,用于针对第一数据集构建第一核密度估计器,所述第一核密度估计器含有值大于0的窗口宽度参数h;
    确定模块,用于确定求取所述窗口宽度参数h的最优值的目标函数,所述目标函数以第二核密度估计器为对数函数的变量,所述第二核密度估计器针对第二数据集构建,所述第二数据集为从所述第一数据集剔除任意一个数据构成;
    求解模块,用于由所述第二核密度估计器和目标函数求解所述窗口宽度参数h的表达式;
    迭代模块,用于选取预设误差阈值ξ和所述窗口宽度参数h的初始值h 0,采用迭代算法将所述窗口宽度参数h的值迭代到与前一次迭代所得h的值的绝对差值不大于所述ξ为止。
  7. 如权利要求6所述的最小熵核密度估计器生成方法,其特征在于,所述构建模块包括:
    估计器构建单元,用于以D={x i|x i∈R,i=1,2,...,L,...,N}为所述第一数据集,构建如下形式的核密度估计器
    Figure PCTCN2018105174-appb-100018
    Figure PCTCN2018105174-appb-100019
    所述R表示实数集,所述N表示所述第一数据集的数据个数,所述h表示所述窗口宽度参数,所述
    Figure PCTCN2018105174-appb-100020
    是高斯核函数
    Figure PCTCN2018105174-appb-100021
  8. 如权利要求6或7所述的最小熵核密度估计器生成方法,其特征在于,所述窗口宽度参数h的最优值的目标函数为:
    Figure PCTCN2018105174-appb-100022
    其中,
    Figure PCTCN2018105174-appb-100023
    为针对所述第二数据集D -i=D-x i构建的第二核密度估计器,
    Figure PCTCN2018105174-appb-100024
    为数据x i对应的概率密度值。
  9. 一种计算设备,包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序,其特征在于,所述处理器执行所述计算机程序时实现如权利要求1至5任意一项所述方法的步骤。
  10. 一种计算机可读存储介质,所述计算机可读存储介质存储有计算机程序,其特征在于,所述计算机程序被处理器执行时实现如权利要求1至5任意一项所述方法的步骤。
PCT/CN2018/105174 2018-09-12 2018-09-12 最小熵核密度估计器生成方法、装置和计算机可读存储介质 WO2020051789A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/105174 WO2020051789A1 (zh) 2018-09-12 2018-09-12 最小熵核密度估计器生成方法、装置和计算机可读存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/105174 WO2020051789A1 (zh) 2018-09-12 2018-09-12 最小熵核密度估计器生成方法、装置和计算机可读存储介质

Publications (1)

Publication Number Publication Date
WO2020051789A1 true WO2020051789A1 (zh) 2020-03-19

Family

ID=69776463

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/105174 WO2020051789A1 (zh) 2018-09-12 2018-09-12 最小熵核密度估计器生成方法、装置和计算机可读存储介质

Country Status (1)

Country Link
WO (1) WO2020051789A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101221238A (zh) * 2008-01-24 2008-07-16 上海交通大学 基于高斯均值移动配准的动态偏差估计方法
US20100100337A1 (en) * 2008-10-16 2010-04-22 Oxfordian, Llc Method to extract parameters from in-situ monitored signals for prognostices
CN102033994A (zh) * 2010-12-07 2011-04-27 北京航空航天大学 基于马尔可夫链蒙特卡罗的舵机可靠性仿真抽样方法
CN107505519A (zh) * 2017-07-26 2017-12-22 中国电力科学研究院 一种分布式电源接入电网电能质量分析方法及装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101221238A (zh) * 2008-01-24 2008-07-16 上海交通大学 基于高斯均值移动配准的动态偏差估计方法
US20100100337A1 (en) * 2008-10-16 2010-04-22 Oxfordian, Llc Method to extract parameters from in-situ monitored signals for prognostices
CN102033994A (zh) * 2010-12-07 2011-04-27 北京航空航天大学 基于马尔可夫链蒙特卡罗的舵机可靠性仿真抽样方法
CN107505519A (zh) * 2017-07-26 2017-12-22 中国电力科学研究院 一种分布式电源接入电网电能质量分析方法及装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
LYAOYUANDEMENG: "L7: Kernel Density Estimation", BAIDU, 14 November 2012 (2012-11-14), pages 9, 19, Retrieved from the Internet <URL:https://wenku.baidu.com/view/befb2d3f376baf1ffc4fad3a.html> *

Similar Documents

Publication Publication Date Title
US20200218509A1 (en) Multiplication Circuit, System on Chip, and Electronic Device
CN107832062B (zh) 一种程序更新方法及终端设备
CN109754359B (zh) 一种应用于卷积神经网络的池化处理的方法及系统
US20220083857A1 (en) Convolutional neural network operation method and device
WO2020119188A1 (zh) 一种程序检测方法、装置、设备及可读存储介质
WO2021000745A1 (zh) 一种知识图谱的嵌入表示方法及相关设备
WO2021051556A1 (zh) 深度学习权值更新方法、系统、计算机设备及存储介质
CN112768056A (zh) 基于联合学习框架的疾病预测模型建立方法和装置
WO2023096571A2 (en) Data processing for release while protecting individual privacy
Liu et al. Task-oriented convex bilevel optimization with latent feasibility
CN108509179B (zh) 用于检测人脸的方法、用于生成模型的装置
CN113965313A (zh) 基于同态加密的模型训练方法、装置、设备以及存储介质
CN116467896B (zh) 一种口腔正畸疗效模拟系统及方法
WO2021109867A1 (zh) 图像处理方法及装置、计算机可读介质和电子设备
WO2020051789A1 (zh) 最小熵核密度估计器生成方法、装置和计算机可读存储介质
WO2021135572A1 (zh) 神经网络的卷积实现方法、卷积实现装置及终端设备
CN112966054A (zh) 基于企业图谱节点间关系的族群划分方法和计算机设备
CN114758130B (zh) 图像处理及模型训练方法、装置、设备和存储介质
US11971796B2 (en) Goal seek analysis based on status models
KR101321840B1 (ko) 퍼지 기반 레티넥스를 이용한 영상 정규화 방법 및 장치
WO2023010701A1 (en) Image generation method, apparatus, and electronic device
CN114356235A (zh) 一种数据标准化处理方法、装置、电子设备及存储介质
CN115640517A (zh) 多方协同模型训练方法、装置、设备和介质
WO2022178975A1 (zh) 基于噪声场的图像降噪方法、装置、设备及存储介质
CN112329715A (zh) 一种人脸识别方法、装置、设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18933576

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 18/06/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 18933576

Country of ref document: EP

Kind code of ref document: A1