Cloud computing analysis method based on time mark pulse sequence

Abstract

The invention discloses a cloud computing analysis method based on a time mark pulse sequence, and belongs to the technical field of signal processing. The method is used for solving the technical problems that in an existing scheme, delay nonlinearity caused by multipath propagation is not considered, and effective decoupling cannot be achieved due to a coupling effect existing in atomic clock error estimation and system deviation. A direct component and a scattering component can be distinguished through time scale layered mapping, and the amplitude-phase-frequency coupling effect of multi-path propagation is covered by fusing the three-dimensional characteristics of delay inequality, amplitude attenuation and frequency drift; the nonlinear mapping capability of the hyperbolic tangent kernel function is superior to that of a linear model, and the hyperbolic tangent kernel function can adapt to time delay nonlinear change in a complex electromagnetic environment; clock error, time delay parameters and system deviation are modeled in a unified mode, error accumulation caused by traditional separation estimation is avoided, variable dependency is quantized through an information theory tool, mutual information decoupling is achieved, robustness is enhanced, and a decoupling result directly corresponds to an atomic clock core index and system interference.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of signal processing, in particular to a cloud computing analysis method based on time-stamped pulse sequence. BACKGROUND

[0002] The computational analysis of pulse sequence is a common problem in the fields of signal processing, neuroscience, time series analysis, communication systems, etc., whose core goal is to measure the difference between two pulse (or event) sequences with precise timestamps; since the pulse sequence is essentially a set of time points of discrete events, the traditional point-by-point difference-based error is not applicable, such as mean square error MSE. Therefore, a specially designed measurement method is needed.

[0003] The time base jitter of the pulse signal source will cause the magnetic signal time reference to be misaligned, resulting in the distortion of the measurement pattern, and the time base jitter of 15ps can increase the main lobe width of the pattern by 23% and raise the side lobe level by 5dB; although the existing technical solution can compensate for the phase error through a reference channel monitoring and a Fourier transform model, this method requires very high hardware synchronization accuracy, and does not consider the time delay nonlinearity caused by multipath propagation; in addition, in the pulsar timekeeping system, the atomic clock error estimation is coupled with the system deviation, and the traditional single-step Kalman filter cannot effectively decouple, resulting in the accumulation of small deviations over time, which seriously affects the long-term timekeeping accuracy. SUMMARY

[0004] The present application relates to the technical field of signal processing, in particular to a cloud computing analysis method based on time-stamped pulse sequence.

[0005] The present application relates to the technical field of signal processing, in particular to a cloud computing analysis method based on time-stamped pulse sequence. The cloud computing analysis method based on time-stamped pulse sequence comprises: time-stamped hierarchical mapping of the received time-stamped pulse sequence, extraction of the time delay nonlinearity feature of multipath propagation, and input of the pre-constructed time delay compensation model based on the hyperbolic tangent kernel function, output of the nonlinear time delay parameter; Based on the output nonlinear time delay parameter, the joint state space equation of the atomic clock error and the system deviation is constructed, and the bivariate mutual information maximization algorithm is used to decouple the clock error estimate and the system deviation; According to the decoupled clock error estimate and system deviation, the output result of the time delay compensation model is combined, and the comprehensive error of the pulse sequence is calculated by the residual iteration compression algorithm; wherein the time delay nonlinearity correction term is dynamically introduced in the iteration process.

[0006] Preferably, a band time-tagged pulse sequence of external sensor output is received, and the pulse sequence is divided into a direct component time-tagged set and a scattered component time-tagged set based on propagation path characteristics.

[0007] Preferably, when extracting the time delay nonlinear characteristic of multipath propagation based on the layered time-tagged set, the time delay offset of the scattered component relative to the direct component, the scattered component amplitude decay rate are calculated, and the instantaneous frequency offset of each component is extracted through wavelet ridge line transformation. The calculated time delay offset, scattered component amplitude decay rate and instantaneous frequency offset are combined in sequence to construct a feature matrix.

[0008] Preferably, a hyperbolic tangent kernel function-based time delay compensation model is constructed with the feature matrix as input, and nonlinear time delay parameters are output.

[0009] Preferably, a hyperbolic tangent kernel function is used to map the feature space. When training the model, the derivative of the objective function is taken and set to zero, and the normal equation set is obtained by simplifying. The nonlinear time delay parameters are solved by matrix inversion of the normal equation set.

[0010] Preferably, when verifying the rationality of the nonlinear time delay parameters, residual inspection and parameter range constraint are included.

[0011] Preferably, the joint state space equation includes a state equation and an observation equation. The state equation describes the dynamic evolution of the state variable, and the corresponding expression is: ; wherein, is the joint state vector at the time; is the joint state vector at the time; is the state transition matrix; is the time index; is the process noise vector; The observation equation relates the state variable to the observation value, and the corresponding expression is: ; wherein, is the observation value vector; is the observation matrix; is the observation noise vector.

[0012] Preferably, when decoupling the clock bias estimate and the system bias using a bivariate mutual information maximization algorithm, a Gaussian kernel function is used to estimate the probability density, and when performing the mutual information maximization iteration, the decoupling matrix is initialized, the 6-dimensional state vector is projected into a 2-dimensional subspace, and the decoupling matrix is iteratively updated to maximize the bivariate mutual information until convergence.

[0013] Preferably, the nonlinear correction term is calculated based on the delay error sequence, and the residual is updated; the delay nonlinear correction term is defined as: ; wherein, is the original delay error of the first sample; is the delay nonlinear reference threshold; is the iteration number; is the maximum iteration number; is the nonlinear curvature parameter; tanh() is the hyperbolic tangent function.

[0014] Preferably, when the iteration convergence judgment and the comprehensive error output are performed, the residual change is calculated and analyzed, and if the analysis result meets the termination requirement, the comprehensive error is output.

[0015] Compared with the prior art, the present application has the following beneficial effects: The present application can distinguish between direct components and scattered components through time scale layer mapping, achieve accurate separation of multipath components, and solve the problem of time delay estimation deviation caused by multipath signal aliasing in traditional methods; by fusing three-dimensional features of delay difference, amplitude attenuation and frequency drift, comprehensive capture of nonlinear features is achieved, which can cover the amplitude-phase-frequency coupling effect of multipath propagation; the nonlinear mapping capability of the hyperbolic tangent kernel function is better than that of the linear model, and can adapt to the nonlinear change of time delay in complex electromagnetic environment, such as ionospheric scintillation and ground reflection; the output nonlinear time delay parameter directly corresponds to the key physical characteristics of multipath propagation, facilitating subsequent system deviation decoupling and error correction.

[0016] The present application models the clock difference, time delay parameter and system deviation uniformly, avoids error accumulation caused by traditional separation estimation, and can effectively reduce the clock difference estimation error; by quantifying the variable dependence relationship through information theory tools, mutual information decoupling is achieved to enhance robustness, and compared with traditional Kalman filter decoupling, the system deviation separation accuracy can be effectively improved; the decoupling result directly corresponds to the core indicators of atomic clocks and system interference, facilitating error tracing and compensation in engineering applications.

[0017] The dynamic correction term in the present application can effectively suppress the multipath time delay nonlinearity, enhance the robustness, and reduce the error fluctuation range; by combining compressive sensing with adaptive convergence, the calculation time can be reduced, and the real-time requirement can be further met. BRIEF DESCRIPTION OF DRAWINGS

[0018] The present application will be further described below with reference to the accompanying drawings.

[0019] Figure 1 is the flow chart of the cloud computing analysis method based on the time scale pulse sequence of the present application. DETAILED DESCRIPTION

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] like Figure 1 As shown, this invention is a cloud computing analysis method based on time-stamped pulse sequences, comprising: The received time-stamped pulse sequence undergoes time-stamp layer mapping to extract the nonlinear time delay features of multipath propagation. These features are then input into a pre-constructed time delay compensation model based on the hyperbolic tangent kernel function, outputting nonlinear time delay parameters. Specific steps include: The received time-stamped pulse sequence from an external sensor is denoted as... ,in, Let be the arrival time stamp of the i-th pulse, in ns; The pulse amplitude is expressed in mV; i is the pulse index, i=1,2,3,…,N, where N is the total number of pulses; external sensor, such as a pulsar detector or atomic clock receiver; Based on the propagation path characteristics, the pulse sequence is divided into a time-stamped set of direct-path components. and scattering component time scale set , Let dir be the arrival time stamp of the k-th direct component pulse, dir represents the direct path, which refers to the path through which the signal propagates directly from the transmitter to the receiver without reflection or scattering, and k is the index of the direct component pulse, k=1,2,...,K,K≤N; Let scat be the arrival time stamp of the m-th scattered component pulse, scat represent the scattering path, which refers to the path of the signal after being reflected / scattered by the atmosphere, ionosphere or surface obstacles to the receiving device, and m is the index of the scattered component pulse, m=1,2,...,M, M≤N; The specific steps for division include: Set initial threshold Based on the typical multipath delay range of pulsar signals, the time difference between adjacent pulse timescales is calculated. ; These are the arrival times of the i-th pulse and the (i-1)-th pulse, respectively. When the time difference is less than the initial threshold, it is determined to be a scattering component and is assigned to the scattering component time scale set; otherwise, it is assigned to the direct component time scale set. The time-scaled set of the scattering components is divided into two layers according to pulse amplitude. Arrange them in descending order, taking the first 30% as strong scattering components and the last 70% as weak scattering components, and label them accordingly. and ; It should be explained that the preprocessed hierarchical time-stamped set provides structured data input for subsequent feature extraction, ensuring targeted analysis of multi-path components.

[0022] Based on the layered time-stamped set, when extracting the time delay nonlinearity characteristics of multipath propagation, the time delay offset of the scattering component relative to the direct component is calculated. ; ; This represents the time delay offset of the m-th scattering component relative to its corresponding direct component; Let m be the m-th timescale in the timescale set of scattering components. It is the k-th timescale in the direct component timescale set; Calculate the amplitude attenuation rate of the scattered component ;in, Let be the amplitude attenuation rate of the m-th scattering component; The amplitude of the scattering component is the intensity of the m-th scattering pulse. The amplitude of the direct component, i.e., the intensity of the k-th direct pulse; Instantaneous frequency shifts of each component are extracted using wavelet ridge transform. The expression involved is: ;in, This represents the instantaneous frequency offset of the m-th scattering component; These are the wavelet transform coefficients. Here, is the wavelet basis function, and s is the scaling factor. It is a phase angle function; It is the partial derivative with respect to time t; The calculated time delay offset Scattering component amplitude attenuation rate Instantaneous frequency shift Combine sequentially to construct the feature matrix. ; It should be explained that the three features are used together as input to the hyperbolic tangent kernel function delay compensation model. Through the nonlinear mapping capability of the kernel function, the high-dimensional features are compressed into low-dimensional delay parameters, ultimately achieving accurate compensation for multi-path delay nonlinearity.

[0023] With characteristic matrix Using the hyperbolic tangent kernel function as input, a time delay compensation model is constructed. Output nonlinear time delay parameters ; The hyperbolic tangent kernel function is used to map the feature space, and the formula is as follows: ;in, The reference feature matrix is ​​generated from a calibration pulse sequence free from multipath interference; This is the kernel function slope parameter, which controls the degree of stretching of the feature space. The default value is 0.5. This is a bias parameter that adjusts the nonlinearity of the kernel function; the default value is -0.2. During model training, the time delay parameter θ is solved using least squares optimization, and the objective function is: ;in, This is the measured time delay difference vector; This is the regularization coefficient, used to prevent overfitting; the default value is [value to be filled in]. ; The square of the L2 norm; is the hyperbolic tangent kernel function matrix with dimension M×3, where M is the total number of scattering components; Taking the derivative of the objective function and setting it to zero, we obtain the normal equation system: Where I is a 3×3 identity matrix; for Transpose of; Solve for θ using matrix inversion of the normal equation system: ; The nonlinear time delay parameters obtained by the solution Mapped to a specific physical meaning, among which, The time delay difference feature weight reflects the contribution of time delay offset to nonlinear error. The larger the weight, the more significant the impact of time delay difference on the final error. The amplitude attenuation characteristic weight reflects the modulation effect of scattering energy loss on time delay nonlinearity; The frequency drift characteristic weight reflects the impact of the Doppler effect or medium dispersion on time delay in multipath propagation; Verifying the rationality of the nonlinear time delay parameter θ includes residual testing and parameter range constraints. When performing residual testing, the fitted residuals are calculated. Analysis to determine whether the fitting residual is <10 ;10 Set according to the pulsar signal delay accuracy requirements; When setting parameter range constraints, ensure All values ​​belong to [0,1]. If the value is outside the range, the regularization coefficient is readjusted and the matrix inversion is performed again. It should be explained that the output nonlinear time delay parameter can accurately quantify the nonlinear time delay characteristics of multipath propagation, and can provide the core mathematical basis for error compensation in pulsar timekeeping systems or atomic clock synchronization.

[0024] In this embodiment of the invention, by using time-scaled layered mapping, direct components and scattered components can be distinguished, achieving accurate separation of multipath components and solving the problem of time delay estimation deviation caused by multipath signal aliasing in traditional methods. By fusing three-dimensional features of time delay difference, amplitude attenuation, and frequency drift, comprehensive capture of nonlinear features is achieved, covering the amplitude-phase-frequency coupling effect of multipath propagation. The nonlinear mapping capability of the hyperbolic tangent kernel function is superior to that of the linear model, and it can adapt to nonlinear time delay changes in complex electromagnetic environments, such as ionospheric scintillation and surface reflection. The output nonlinear time delay parameters directly correspond to the key physical characteristics of multipath propagation, facilitating subsequent system deviation decoupling and error correction.

[0025] Based on the output nonlinear time delay parameters, a joint state-space equation for the atomic clock error and system bias is constructed, and a bivariate mutual information maximization algorithm is used to decouple the clock error estimate and the system bias. The specific steps include: When constructing the joint state-space equations of atomic clock error and system bias, nonlinear time delay parameters are used. Define a joint state vector that includes atomic clock bias and system bias. , For the first The atomic clock error at a given moment, measured in nanoseconds (ns), represents the deviation between the actual time of the atomic clock and the standard time. For the first The clock drift rate at any given time, expressed in ns / s, reflects the rate of change of clock bias over time. For the first The overall system deviation at any given time, expressed in nanoseconds, includes non-time-delay errors such as equipment noise and environmental interference. For time indexing; The joint state-space equations include state equations and observation equations; The state equation describes the dynamic evolution of state variables, and its corresponding expression is: ;in, This is the state transition matrix; The process noise vector follows a Gaussian distribution. , The process noise covariance matrix; The observation equation, which correlates the state variables with the observed values, is expressed as follows: ;in, The observation matrix; The observed noise vector follows a Gaussian distribution. , To observe the noise covariance matrix; This is the observation vector, i.e., the total measured time delay error. ; It should be explained that the state equation and the observation equation constitute a complete dynamic system model. The influence of nonlinear time delay parameters is explicitly incorporated into the observation process, which can provide a mathematical basis for subsequent decoupling.

[0026] When initializing system parameters, the initial value of the joint state vector is... ,in, The initial values ​​for the nonlinear time delay parameters of the least squares optimization output; Process noise covariance ; Observation noise covariance ; When using the bivariate mutual information maximization algorithm to decouple clock error estimates from system bias, the bivariate mutual information... The calculation formula is: ;in, This is an estimate of the atomic clock bias. , which is the systematic deviation; Let X be the joint probability density function of Y. The marginal probability density function; The probability density is estimated using the Gaussian kernel function, and the relevant expression is: ;in, The probability density function is the estimated value, representing the probability density at the estimated point x, reflecting the likelihood of the random variable taking values ​​near x; x is the independent variable to be estimated, i.e., the target point for probability density estimation. Let j be the j-th sample data point; n be the number of samples; For core bandwidth; It is the normalization constant for the normal distribution; It is an exponential function; When performing the maximum mutual information iteration, initialize the decoupling matrix. Project the 6-dimensional state vector onto a 2-dimensional subspace; Iterative updates to W ensure bivariate mutual information. Maximize until convergence: ; The decoupling result is: ; ; These are the row vectors of the optimal decoupling matrix; It needs to be explained that by maximizing mutual information, the clock error and system bias in the high-dimensional state vector are projected to the low-dimensional space and separated. Information theory tools are used to quantify the dependencies between variables, ensuring that the physical meaning of decoupling is clear.

[0027] When verifying and outputting the decoupling results, calculate the correlation coefficient between the clock error and the system deviation after decoupling. For the correlation coefficient Perform data analysis to determine whether it meets the requirements. The requirement is <0.1; the relevant expression is: ;in, This is the arithmetic mean of the sequence of clock error estimates; The numerator is the arithmetic mean of the series of system bias estimates; the numerator is the covariance calculation term; the denominator is the standard deviation product term. If the requirements are met, output the clock bias estimate. Deviation from the system ; Otherwise, readjust the core bandwidth and repeat the above steps.

[0028] In this embodiment of the invention, clock bias, time delay parameters and system deviation are modeled in a unified manner to avoid error accumulation caused by traditional separate estimation, which can effectively reduce clock bias estimation error; by quantifying variable dependencies through information theory tools, mutual information decoupling is achieved to enhance robustness, which can effectively improve the system deviation separation accuracy compared with traditional Kalman filter decoupling; the decoupling results directly correspond to the core indicators of atomic clock and system interference, which facilitates error tracing and compensation in engineering applications.

[0029] Based on the decoupled clock error estimate and system deviation, combined with the output of the time delay compensation model, the comprehensive error of the pulse sequence is calculated using a residual iterative compression algorithm; wherein, a time delay nonlinearity correction term is dynamically introduced during the iteration process. The specific steps include: Based on the decoupled clock error estimates and system bias, obtain the corresponding atomic clock error estimate sequence. and system bias estimate sequence ; The total number of samples; And, obtain the time delay error sequence output by the time delay compensation model. ; When initializing parameters, include the initial synthesis error vector. ; Iterative residual Number of iterations Maximum number of iterations Convergence threshold ; Time delay nonlinearity correction term coefficient Nonlinear curvature parameters ; Calculate the nonlinear correction term based on the time delay error sequence. And update the residuals, involving the following expressions: ;in, To correct the residual vector; For the first The residual vector of the next iteration; The time delay nonlinearity correction term is defined as follows: ;in, Let be the time delay nonlinear correction term, representing the th In the iteration, for the first Time delay error per sample The amount of nonlinear compensation performed; For the first The original time delay error of each sample; This serves as a time-delay nonlinear reference threshold. This represents the number of iterations. The maximum number of iterations is given; tanh() is the hyperbolic tangent function. When performing residual iterative compression and sparse optimization, a sparse observation matrix is ​​constructed. By defining dimension 100× , For the signal dimension, generate Gaussian random numbers with a mean of 0 and a variance of 1 / 100, fill the matrix with the Gaussian random numbers and select column normalization to obtain the result; Minimizing the compressed residual using the L1 norm involves the following expression: ;in, To optimize the residual vector, let represent the solution obtained by minimizing the L1 norm. The residual vector of the next iteration; For minimization operators, let represent the optimal solution for finding the minimum value of the objective function with respect to variable r; The term is the squared term of the L2 norm; Here is the sparse regularization parameter, and its value range is... ; The L1 norm term represents the sum of the absolute values ​​of all elements in the residual vector r; When performing iterative convergence judgment and comprehensive error output, calculate the residual change. , For the first The residual vector of the next iteration; the change in residual. Perform analysis; like ,or If the iteration terminates, the combined error is output. ; Otherwise, let And repeat the above steps.

[0030] In this embodiment of the invention, the dynamic correction term can effectively suppress multipath delay nonlinearity, enhance robustness, and reduce the error fluctuation range; by combining compressed sensing with adaptive convergence, computation time can be reduced, and real-time requirements can be further met.

[0031] In the several embodiments provided by this invention, it should be understood that the disclosed system can be implemented in other ways. For example, the embodiments of the invention described above are merely illustrative; for example, the division of modules is only a logical functional division, and there may be other division methods in actual implementation.

[0032] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0033] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or in the form of hardware plus software functional modules.

[0034] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the essential characteristics of the present invention.

[0035] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A cloud computing analysis method based on time-scaled pulse sequences, characterized in that, include: The received time-stamped pulse sequence is subjected to time-stamp layer mapping to extract the nonlinear time delay features of multipath propagation. The model is then input into a pre-constructed time delay compensation model based on the hyperbolic tangent kernel function, and the nonlinear time delay parameters are output. Based on the output nonlinear time delay parameters, a joint state-space equation for atomic clock error and system deviation is constructed, and a bivariate mutual information maximization algorithm is used to decouple the clock error estimate and the system deviation. Based on the decoupled clock error estimate and system deviation, combined with the output of the time delay compensation model, the comprehensive error of the pulse sequence is calculated using the residual iterative compression algorithm; wherein, a time delay nonlinear correction term is dynamically introduced during the iteration process.

2. The cloud computing analysis method based on time-stamped pulse sequences according to claim 1, characterized in that, The system receives time-stamped pulse sequences from external sensors and, based on propagation path characteristics, divides the pulse sequences into a direct component time-stamp set and a scattering component time-stamp set.

3. The cloud computing analysis method based on time-stamped pulse sequences according to claim 2, characterized in that, Based on the layered time-stamped set, when extracting the time delay nonlinearity features of multipath propagation, the time delay offset of the scattering component relative to the direct component and the amplitude attenuation rate of the scattering component are calculated, and the instantaneous frequency offset of each component is extracted by wavelet ridge transform. The calculated time delay offset, scattering component amplitude attenuation rate, and instantaneous frequency offset are combined in sequence to construct a feature matrix.

4. The cloud computing analysis method based on time-stamped pulse sequences according to claim 3, characterized in that, Using the feature matrix as input, a time delay compensation model based on the hyperbolic tangent kernel function is constructed, and nonlinear time delay parameters are output.

5. The cloud computing analysis method based on time-stamped pulse sequences according to claim 4, characterized in that, The feature space is mapped using the hyperbolic tangent kernel function; When training the model, the derivative of the objective function is taken and set to zero, which simplifies to obtain the normal equation system. The nonlinear time delay parameters are solved by matrix inversion of the normal equation system.

6. The cloud computing analysis method based on time-stamped pulse sequences according to claim 5, characterized in that, Verifying the rationality of nonlinear time delay parameters includes residual testing and parameter range constraints.

7. The cloud computing analysis method based on time-stamped pulse sequences according to claim 6, characterized in that, The joint state-space equations include state equations and observation equations; The state equation describes the dynamic evolution of state variables, and its corresponding expression is: ;in, For the first The joint state vector at time t; For the first The joint state vector at time t; This is the state transition matrix; For time indexing; This is the process noise vector; The observation equation, which correlates the state variables with the observed values, is expressed as follows: ;in, A vector of observations; The observation matrix; This is the observed noise vector.

8. The cloud computing analysis method based on time-stamped pulse sequences according to claim 7, characterized in that, When using the bivariate mutual information maximization algorithm to decouple the clock error estimate from the system deviation, the probability density is estimated using the Gaussian kernel function. When performing the mutual information maximization iteration, the decoupling matrix is ​​initialized and the 6-dimensional state vector is projected onto the 2-dimensional subspace. The decoupling matrix is ​​iteratively updated to maximize the mutual information between the two variables until convergence.

9. The cloud computing analysis method based on time-stamped pulse sequences according to claim 8, characterized in that, The nonlinear correction term is calculated based on the time delay error sequence, and the residual is updated; the time delay nonlinear correction term Defined as: ;in, For the first The original time delay error of each sample; This serves as a time-delay nonlinear reference threshold. This represents the number of iterations. This represents the maximum number of iterations. is the nonlinear curvature parameter; tanh() is the hyperbolic tangent function.

10. The cloud computing analysis method based on time-stamped pulse sequences according to claim 9, characterized in that, When performing iterative convergence judgment and comprehensive error output, the residual change is calculated and analyzed. If the analysis result meets the termination requirements, the comprehensive error is output.

Images