Method, device, equipment and medium for improving transient learning performance of working machine

Abstract

The application relates to the field of machine learning, and discloses a method, device, equipment and medium for improving transient learning performance of a worker. The method comprises the following steps: constructing a training set for local prediction of each worker, training each worker on the training set, and obtaining a variance of local prediction of each worker on test data; aggregating the variance of local prediction by a server to obtain a variance of global prediction on the test data; and setting a fusion algorithm for each worker according to the variance of global prediction and the variance of local prediction to obtain an optimal prediction model of each worker. The method disclosed by the application can ensure that the transient performance of each worker is improved after local prediction and global prediction are fused.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of machine learning, and in particular to a method and device for improving transient learning performance of a working machine, and a medium. BACKGROUND

[0002] The Internet of Things generates a large amount of distributed data. A typical training method is to store the data on a server and train the model through the server. However, the communication efficiency and the computing efficiency of this method are obviously problematic. For example, the hundreds of Gb of data generated by a car in a few hours is a huge burden in the transmission and computing process. Distributed machine learning can effectively solve the problem of too much computing, too much training data, and too large a model size. In practical applications, distributed machine learning generally uses a deep neural network as a machine learning model. This machine learning model has achieved unprecedented success in many applications, such as model classification and pattern recognition, but is mainly limited to offline learning. In practical applications, a working machine obtains a data stream, so online learning is an effective way to solve this problem.

[0003] In the prior art, after the working machine completes local prediction and the server completes global prediction, the fused prediction error and the local prediction error are directly compared, which is not easy to prove in theory. In addition, the concept of mean square error is used in the prior art, and no good results have been achieved, especially in the transient response of the working machine, and the learning performance improvement result is not good. SUMMARY

[0004] Therefore, the present application proposes a method and device for improving the transient learning performance of a working machine, and a medium. The method for improving the transient learning performance of the working machine proposed by the present application uses Gaussian process regression (GPR). Gaussian process regression can process streaming data, predict test output, and send the expectation and variance of local prediction to the server. The server aggregates the expectations and variances of all working machines received to obtain the global prediction expectation and variance, and sends the global prediction expectation and variance back to each working machine. The optimal prediction model of the working machine is constructed by comparing the variance of local prediction and the variance of global prediction. If the variance of global prediction is not greater than the variance of local prediction, the variance of local prediction is replaced with the variance of global prediction; if the variance of global prediction is greater than the variance of local prediction, the variance of local prediction remains unchanged. At the same time, an indirect method can be used for verification. The upper bound of the fused prediction error is smaller than the upper bound of the original local prediction error, that is, the fused prediction value has a smaller confidence interval.

[0005] To achieve the above object, one aspect of an embodiment of the present application provides a method for improving transient learning performance of a worker, the method comprising the steps of: constructing a training set for local prediction by each worker, training the each worker on the training set, and obtaining a variance of local prediction of test data by the each worker; aggregating the variance of local prediction by a server to obtain a variance of global prediction of test data; and setting a fusion algorithm for the each worker according to the variance of global prediction and the variance of local prediction to obtain an optimal prediction model of the each worker.

[0006] In some embodiments, the step of constructing a training set for local prediction by each worker, training the each worker on the training set, and obtaining a variance of local prediction of test data by the each worker comprises the steps of: constructing an objective function and constructing a training set based on the objective function, calculating a projection of test data into the training set to obtain projection data corresponding to the test data; constructing a new training set according to the projection data corresponding to the test data, and calculating a Gaussian posterior probability distribution on the new training set to obtain the variance of local prediction of test data by the each worker.

[0007] In some embodiments, the step of aggregating the variance of local prediction by a server to obtain a variance of global prediction of test data comprises the steps of: sending the variance of local prediction of test data by the each worker to the server, and aggregating the variance of local prediction by an aggregation algorithm on the server to obtain the variance of global prediction of test data.

[0008] In some embodiments, the step of setting a fusion algorithm for the each worker according to the variance of global prediction and the variance of local prediction to obtain an optimal prediction model of the each worker comprises the steps of:

[0009] sending the variance of global prediction to the each worker and constructing a set of test data of uncertainty, and setting a fusion algorithm according to the variance of global prediction and the variance of local prediction of the each worker on the set of test data of uncertainty to obtain the optimal prediction model of the each worker.

[0010] In some embodiments, the sending the global predicted variance to each worker and constructing a set of test data of uncertainty, setting a fusion algorithm on the set of test data of uncertainty according to the global predicted variance and the local predicted variance of each worker, and obtaining the optimal prediction model of each worker further comprises: comparing the global predicted variance and the local predicted variance of each worker on the set of test data of uncertainty; and in response to the global predicted variance being not greater than the local predicted variance of each worker, setting the optimal prediction model of each worker by the global prediction.

[0011] In some embodiments, the sending the global predicted variance to each worker and constructing a set of test data of uncertainty, setting a fusion algorithm on the set of test data of uncertainty according to the global predicted variance and the local predicted variance of each worker, and obtaining the optimal prediction model of each worker further comprises: in response to the global predicted variance being greater than the local predicted variance of each worker, setting the optimal prediction model of each worker by the local prediction of each worker.

[0012] In some embodiments, the method further comprises: verifying whether the prediction performance of the optimal prediction model is optimal prediction performance according to the size of the error of the local prediction and the error of the global prediction.

[0013] Another aspect of the embodiments of the present application also provides a device for improving transient learning performance of a worker, the device comprising: a first module configured to construct a training set for local prediction of each worker, train each worker on the training set, and obtain a local predicted variance of each worker for test data; a second module configured to aggregate the local predicted variance by a server to obtain an expected variance of global prediction for test data; and a third module configured to set a fusion algorithm for each worker according to the global predicted variance and the local predicted variance, and obtain an optimal prediction model of each worker.

[0014] In some embodiments, the first module is further configured to: construct an objective function and construct a training set based on the objective function, calculate a projection of test data into the training set to obtain corresponding projection data of the test data; construct a new training set according to the corresponding projection data of the test data, calculate a Gaussian posterior probability distribution on the new training set to obtain the local predicted variance of each worker for test data.

[0015] In some embodiments, the second module is further configured to: send the variance of the local predictions of the test data by each worker to the server, and aggregate the variance of the local predictions on the server using an aggregation algorithm to obtain the variance of the global predictions of the test data.

[0016] In some embodiments, the third module is further configured to: send the variance of the global prediction to each worker and construct a test data set with uncertainty, and set a fusion algorithm on the test data set with uncertainty based on the variance of the global prediction and the variance of the local prediction of each worker to obtain the optimal prediction model for each worker.

[0017] In some embodiments, the third module is further configured to: compare the variance of the global prediction with the variance of the local prediction of each worker on the set of uncertain test data; and, in response to the global prediction variance not being greater than the local prediction variance of each worker, set the optimal prediction model for each worker using the global prediction.

[0018] In some embodiments, the third module is further configured to: in response to the global prediction variance being greater than the local prediction variance of each worker, set the optimal prediction model for each worker based on the local prediction of each worker.

[0019] In another aspect of the present invention, a computer device is also provided, including at least one processor; and a memory storing computer instructions executable on the processor, the instructions, when executed by the processor, implementing the steps of any of the methods described above.

[0020] In another aspect of the present invention, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements any of the method steps described above.

[0021] The present invention has at least the following beneficial effects: The present invention proposes a method, apparatus, device and medium for improving the transient learning performance of a working machine. The method for improving the transient learning performance of a working machine proposed in the present invention can theoretically guarantee the transient performance improvement of each working machine after fusing local prediction and global prediction. At the same time, the method of the present invention also verifies the prediction results, demonstrating the effectiveness of the obtained optimal prediction model, so that the theoretical and actual verification results are consistent. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings without creative effort.

[0023] Figure 1 A schematic diagram illustrating an embodiment of a method for improving the transient learning performance of a working machine provided by the present invention;

[0024] Figure 2 A schematic diagram illustrating an embodiment of the local prediction performance of the working machine provided by the present invention;

[0025] Figure 3 A schematic diagram illustrating an embodiment of the prediction performance of the optimal prediction model provided by the present invention;

[0026] Figure 4 A schematic diagram of an embodiment of a device for improving the transient learning performance of a working machine provided by the present invention;

[0027] Figure 5 A schematic diagram illustrating an embodiment of a computer device provided by the present invention;

[0028] Figure 6 This is a schematic diagram of an embodiment of a computer-readable storage medium provided by the present invention. Detailed Implementation

[0029] The following describes embodiments of the present invention. However, it should be understood that the disclosed embodiments are merely examples, and other embodiments may take various alternative forms.

[0030] Furthermore, it should be noted that all uses of the terms "first" and "second" in the embodiments of this invention are for the purpose of distinguishing two entities or parameters with the same name but different names. Therefore, "first" and "second" are merely for convenience of expression and should not be construed as limiting the embodiments of this invention. Subsequent embodiments will not elaborate on this further. The terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements may include not only those elements but also elements not expressly listed or inherent to those processes, methods, articles, or apparatuses.

[0031] One or more embodiments of this application will now be described with reference to the accompanying drawings.

[0032] Based on the above objectives, a first aspect of the present invention provides an embodiment of a method for improving the transient learning performance of a working machine. Figure 1The diagram shown is a schematic representation of an embodiment of a method for improving the transient learning performance of a working machine provided by the present invention. Figure 1 As shown, a method for improving the transient learning performance of a working machine according to an embodiment of the present invention includes the following steps:

[0033] S1. Construct a training set for each worker machine for local prediction, train each worker machine on the training set, and obtain the variance of each worker machine's local prediction of the test data.

[0034] S2. The variance of the local predictions is aggregated by the server to obtain the variance of the global prediction for the test data.

[0035] S3. Based on the variance of the global prediction and the variance of the local prediction, a fusion algorithm is set for each working machine to obtain the optimal prediction model for each working machine.

[0036] According to several embodiments of the present invention, constructing a training set for each working machine for local prediction, and training each working machine on the training set to obtain the variance of each working machine's local prediction of the test data includes: constructing an objective function and constructing a training set based on the objective function; calculating the projection of the test data onto the training set to obtain the projection data corresponding to the test data; constructing a new training set based on the projection data corresponding to the test data; calculating the Gaussian posterior probability distribution on the new training set to obtain the variance of each working machine's local prediction of the test data.

[0037] According to several embodiments of the present invention, the variance of the global prediction of the test data is obtained by aggregating the variance of the local predictions by the server, which includes: sending the variance of the local predictions of the test data of each worker machine to the server, and aggregating the variance of the local predictions by the server through an aggregation algorithm to obtain the variance of the global prediction of the test data.

[0038] According to several embodiments of the present invention, setting a fusion algorithm for each working machine based on the variance of global prediction and the variance of local prediction to obtain the optimal prediction model for each working machine includes: sending the variance of global prediction to each working machine and constructing an uncertain test data set; setting a fusion algorithm on the uncertain test data set based on the variance of global prediction and the variance of local prediction for each working machine to obtain the optimal prediction model for each working machine.

[0039] According to several embodiments of the present invention, the variance of the global prediction is sent to each working machine and an uncertain test data set is constructed. A fusion algorithm is set on the uncertain test data set based on the variance of the global prediction and the variance of the local prediction of each working machine to obtain the optimal prediction model for each working machine. This includes: comparing the variance of the global prediction and the variance of the local prediction of each working machine on the uncertain test data set; and setting the optimal prediction model for each working machine based on the global prediction if the variance of the global prediction is not greater than the variance of the local prediction of each working machine.

[0040] According to several embodiments of the present invention, the variance of the global prediction is sent to each working machine and an uncertain test data set is constructed. On the uncertain test data set, a fusion algorithm is set according to the variance of the global prediction and the variance of the local prediction of each working machine to obtain the optimal prediction model of each working machine. The method further includes: in response to the variance of the global prediction being greater than the variance of the local prediction of each working machine, the optimal prediction model of each working machine is set through the local prediction of each working machine.

[0041] According to several embodiments of the present invention, the method of the present invention may further include: verifying whether the prediction performance of the optimal prediction model is the optimal prediction performance based on the magnitude of the error of the local prediction and the error of the global prediction.

[0042] The following provides specific steps of another embodiment of the method for improving the transient learning performance of the working machine based on the present invention.

[0043] Define the objective function as in is n x The input space is one-dimensional. Without loss of generality, we assume the output is one-dimensional, i.e., At time t, given The corresponding output is

[0044] y=f(x)+ε (1)

[0045] ε follows a pattern with a mean of 0 and a variance of . Gaussian noise with a Gaussian probability distribution, i.e. Define a training set in the following form in The input data set is y = [y(1), y(2), ..., y(n)]. s )] Τ It is a column vector that aggregates the outputs. The goal of Gaussian process regression is to utilize the training set... In the test dataset The function f is approximated.

[0046] Define a symmetric positive semi-definite kernel function. That is, ∫k(x,x')f(x)f(x')dν(x)dν(x')≥0, where ν is a measure. Let Return a column vector such that its i-th element equals f(x(i)). Assume the function f is a sample from a Gaussian process prior probability distribution with mean μ and kernel function k. Then the training output and test output... Follows a joint probability distribution

[0047]

[0048] in and Return the values ​​of μ(x(i)) and μ(x). * (i) is a vector. Return a matrix such that the element in the i-th row and j-th column is k(x(i), x). * (j)).

[0049] Gaussian process regression utilizes the properties of Gaussian processes and the training set. Predict test dataset The output. This output It follows a normal distribution, that is here

[0050]

[0051] In distributed machine learning, consider a network with n worker machines. Define this set as... At every moment t, every working machine Utilizing local training data To predict the function for the test input The output. y i (t)=[y i (1),...,y i The local prediction value for each worker machine during training is (t)

[0052]

[0053] In a federated learning framework, each worker machine will generate its trained local predictions. cov i (f * Send it to the server.

[0054] The following are the specific steps for distributed training and fusion.

[0055] (1) Construct a training subset based on the projection of the training set, and define the distance between two training data points x and x' as d(x,x')=||x-x'||, where x is the distance from the data point x to the set. The distance is Define data point x to set The projection set is

[0056] Consider each working machine and its local training dataset For a test data x * Calculate the test data x * To training set Projection:

[0057]

[0058] At each time t, this set of local projections Includes One projection data, namely

[0059] For each working machine and its projection set Take out each projection point and label it as... Here, the subscript j represents the j-th projection point. Then, for each projection point... Find one of its neighborhoods Make And targeting It's important to note that the number of neighborhoods is adjustable, allowing for a fixed selection.

[0060] For each working machine Construct a new training set

[0061] (2) Selecting a kernel function: In practical applications, a kernel function is generally selected.

[0062]

[0063] (3) For each working machine In the new training set Calculate the Gaussian posterior probability distribution above, i.e.:

[0064]

[0065] In the training subset Local predictions are obtained using formula (7). and cov i (f *Then, this local prediction is sent to the server. It can be proven that the local prediction error is less than an upper bound, which we define as g. i (cov i (f * For the test input, the following inequality holds.

[0066]

[0067] (4) The server uses the following aggregation algorithm to aggregate the local predictions and gives the global prediction expectation and variance:

[0068]

[0069] Then, the server will globally predict the expected value. and variance cov global (f * By sending the data back to each working machine, it can be verified that the global prediction error is less than an upper bound, which depends on the cov value. global (f * We express the prediction error inequality as:

[0070]

[0071] (5) Based on the global prediction variance cov rbcm (f * ) and local prediction variance cov i (f * For each working machine, a fusion algorithm is designed to make the fused prediction more closely approximate the true value of the objective function f. A test dataset x with small uncertainties is constructed. * The set is as follows:

[0072]

[0073] If the set is not empty, the global prediction from the server will be used; if the set is empty, the local prediction from the worker will be used.

[0074] Since it cannot be rigorously proven that the prediction error after fusion is less than the local prediction error, i.e. However, theoretical results can be obtained using indirect methods. If the upper bound of the fused prediction error is g(cov... global (f * The local prediction error upper bound g of each working machine is less than that of the working machine. i (cov i (f *This indicates that the optimal prediction model obtained after fusion has a narrower confidence interval. This indirect method demonstrates that the optimal prediction model obtained after fusion improves the accuracy of the prediction results of each worker machine. Secondly, the discussion focuses on the predicted value at time t for each prediction input point, i.e., the transient performance of the worker machine's prediction. global (f * ) <cov i (f * ), then g(cov) global (f * )) <g i (cov i (f * )).

[0075] (6) Using the constructed dataset, verify the effectiveness of the optimal prediction model obtained after fusion. For formula (1), establish the following objective function:

[0076] f(x) = 5x 2 sin(12x)+(x 3 -0.5)sin(3x-0.5)+4cos(2x) (12)

[0077] and Gaussian noise Generate 10 in the interval [0, 1]. 4 We used the training data and randomly selected 120 test points. The network consisted of 40 worker machines, and the training data was divided into 40 equal parts, giving each worker machine 250 training data points. Figure 2 and Figure 3 The diagram shows the local prediction performance of the working machine and the prediction performance of the optimal prediction model obtained after fusion. Figure 2 and Figure 3 The continuous curve in the graph represents the actual measured output, while the discontinuous dashed line represents the predicted value based on local data. Figure 2 It can be seen that the predicted values ​​fluctuate wildly around the actual measured values, exhibiting drastic oscillations, and the confidence intervals have jagged edges, increasing the uncertainty of the prediction. From Figure 3 As can be seen, the curve and the discontinuous dashed line have a high degree of overlap, and their confidence intervals are narrower, meaning that the prediction accuracy of the optimal prediction model obtained after fusion has reached the expected level. Therefore, we see a significant improvement in the transient performance of each working machine.

[0078] A second aspect of the present invention provides an apparatus for improving the transient learning performance of a working machine. Figure 4 The diagram shown is a schematic representation of an embodiment of a device for improving the transient learning performance of a working machine according to the present invention. Figure 4As shown, the present invention provides an apparatus for improving the transient learning performance of a worker machine, comprising: a first module configured to construct a training set for local prediction of each worker machine, train each worker machine on the training set, and obtain the expectation and variance of the local prediction of each worker machine for test data; a second module configured to aggregate the expectation and variance of the local prediction through a server to obtain the expectation and variance of the global prediction for the test data; and a third module configured to set a fusion algorithm for each worker machine based on the variance of the global prediction and the variance of the local prediction to obtain the optimal prediction model of each worker machine, and verify the optimal prediction model of each worker machine based on the expectation of the global prediction and the expectation of the local prediction.

[0079] To achieve the above objectives, a third aspect of the present invention provides a computer device. Figure 5 The diagram shown is a schematic representation of an embodiment of a computer device provided by the present invention. Figure 5 As shown, an embodiment of a computer device provided by the present invention includes the following modules: at least one processor 021; and a memory 022, the memory 022 storing computer instructions 023 that can be executed on the processor 021, the computer instructions 023 implementing the steps of the method described above when executed by the processor 021.

[0080] The present invention also provides a computer-readable storage medium. Figure 6 The diagram shown is a schematic representation of an embodiment of a computer-readable storage medium provided by the present invention. Figure 6 As shown, computer-readable storage medium 031 stores a computer program 032 that, when executed by a processor, performs the methods described above.

[0081] Finally, it should be noted that those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program for setting system parameters can be stored in a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the above methods. The storage medium for the program can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc. The above computer program embodiments can achieve the same or similar effects as any of the corresponding foregoing method embodiments.

[0082] Furthermore, the method disclosed in the embodiments of the present invention can also be implemented as a computer program executed by a processor, which may be stored in a computer-readable storage medium. When the computer program is executed by the processor, it performs the functions defined in the method disclosed in the embodiments of the present invention.

[0083] Furthermore, the above-described method steps and system units can also be implemented using a controller and a computer-readable storage medium for storing a computer program that enables the controller to perform the functions of the above-described steps or units.

[0084] Those skilled in the art will also understand that the various exemplary logic blocks, modules, circuits, and algorithm steps described in conjunction with the disclosure herein can be implemented as electronic hardware, computer software, or a combination of both. To clearly illustrate this interchangeability between hardware and software, the functionality of various illustrative components, blocks, modules, circuits, and steps has been generally described. Whether this functionality is implemented as software or as hardware depends on the specific application and the design constraints imposed on the system as a whole. Those skilled in the art can implement the functionality in various ways for each specific application, but such implementation decisions should not be construed as departing from the scope of the embodiments disclosed herein.

[0085] In one or more exemplary designs, functionality may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functionality may be stored as one or more instructions or code on or transmitted via a computer-readable medium. Computer-readable media include computer storage media and communication media, including any medium that facilitates the transfer of a computer program from one location to another. Storage media may be any available medium accessible to a general-purpose or special-purpose computer. By way of example, and not limitation, computer-readable media may include RAM, ROM, EEPROM, CD-ROM or other optical disc storage devices, disk storage devices or other magnetic storage devices, or any other medium that may be used to carry or store the required program code in the form of instructions or data structures and is accessible to a general-purpose or special-purpose computer or a general-purpose or special-purpose processor. Furthermore, any connection may be appropriately referred to as computer-readable media. For example, if software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the aforementioned coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are all included in the definition of media. As used herein, disks and optical discs include compact discs (CDs), laser discs, optical discs, digital versatile discs (DVDs), floppy disks, and Blu-ray discs, where disks typically reproduce data magnetically, while optical discs reproduce data optically using lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0086] The above are exemplary embodiments disclosed in this invention. However, it should be noted that various changes and modifications can be made without departing from the scope of the embodiments of this invention as defined by the claims. The functions, steps, and / or actions of the methods according to the disclosed embodiments described herein do not need to be performed in any particular order. Furthermore, although the elements disclosed in the embodiments of this invention may be described or claimed individually, they may be understood as multiple unless explicitly limited to a singular number.

[0087] It should be understood that, as used herein, the singular form “a” is intended to include the plural form as well, unless the context clearly supports an exception. It should also be understood that, as used herein, “and / or” refers to any and all possible combinations of one or more of the associated listed items.

[0088] The embodiment numbers disclosed in the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0089] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0090] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples. Within the framework of the invention, technical features of the above embodiments or different embodiments can be combined, and many other variations of different aspects of the invention exist, which are not provided in the details for the sake of brevity. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.

Claims

1. A method for improving the transient learning performance of a working machine, characterized in that, include: Construct a training set for each worker machine for local prediction, train each worker machine on the training set, and obtain the variance of each worker machine's local prediction of the test data; The variance of the local predictions is aggregated by the server to obtain the variance of the global prediction for the test data. The optimal prediction model for each worker is obtained by setting a fusion algorithm based on the variance of the global prediction and the variance of the local prediction, comprising: sending the variance of the global prediction to each worker and constructing an uncertain test data set; setting a fusion algorithm on the uncertain test data set based on the variance of the global prediction and the variance of the local prediction of each worker to obtain the optimal prediction model for each worker; wherein, the variance of the global prediction and the variance of the local prediction of each worker are compared on the uncertain test data set; in response to the variance of the global prediction not being greater than the variance of the local prediction of each worker, the optimal prediction model for each worker is set based on the global prediction; in response to the variance of the global prediction being greater than the variance of the local prediction of each worker, the optimal prediction model for each worker is set based on the local prediction of each worker. The optimal prediction model's performance is verified by comparing the magnitudes of the local prediction error and the global prediction error.

2. The method according to claim 1, characterized in that, The construction of a training set for each worker machine for local prediction, and the training of each worker machine on the training set to obtain the variance of each worker machine's local prediction of the test data, includes: Construct an objective function and build a training set based on the objective function; calculate the projection of the test data onto the training set to obtain the projection data corresponding to the test data. A new training set is constructed based on the projection data corresponding to the test data. A Gaussian posterior probability distribution is calculated on the new training set to obtain the variance of the local prediction of the test data by each working machine.

3. The method according to claim 1, characterized in that, The step of aggregating the variance of the local predictions through the server to obtain the variance of the global prediction for the test data includes: The variance of each worker's local prediction of the test data is sent to the server, where the variance of the local prediction is aggregated using an aggregation algorithm to obtain the variance of the global prediction of the test data.

4. A device for improving the transient learning performance of a machine tool, characterized in that, The device includes: The first module configures a training set for each worker machine to perform local predictions, trains each worker machine on the training set, and obtains the variance of each worker machine's local predictions of the test data. The second module is configured to aggregate the variance of the local predictions via a server to obtain the expected variance of the global predictions for the test data; and The third module is configured to set a fusion algorithm for each worker based on the variance of the global prediction and the variance of the local prediction to obtain the optimal prediction model for each worker. This includes: sending the variance of the global prediction to each worker and constructing an uncertain test data set; setting a fusion algorithm on the uncertain test data set based on the variance of the global prediction and the variance of the local prediction of each worker to obtain the optimal prediction model for each worker; wherein, the variance of the global prediction and the variance of the local prediction of each worker are compared on the uncertain test data set; in response to the variance of the global prediction not being greater than the variance of the local prediction of each worker, the optimal prediction model for each worker is set based on the global prediction; in response to the variance of the global prediction being greater than the variance of the local prediction of each worker, the optimal prediction model for each worker is set based on the local prediction of each worker. The third module is also configured to: verify the optimal prediction model for each worker based on the expectation of the global prediction and the expectation of the local prediction.

5. A computer device, characterized in that, include: At least one processor; as well as A memory storing computer instructions executable on the processor, which, when executed by the processor, implement the steps of the method according to any one of claims 1-3.

6. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-3.

Images