WO2023166765A1

WO2023166765A1 - Data processing method, apparatus, device, and medium

Info

Publication number: WO2023166765A1
Application number: PCT/JP2022/034138
Authority: WO
Inventors: 九廷陳; 源連; 天奇黄
Original assignee: 恒林日本株式会社
Priority date: 2022-03-04
Filing date: 2022-09-13
Publication date: 2023-09-07
Also published as: JPWO2023166765A1; JP7390088B1; CN114692873A

Abstract

The present invention relates to a data processing method, apparatus, device, and medium. The apparatus comprises: a model processing unit for reading a model processing rule, and processing the model according to the rule; a data processing unit for reading a data processing rule and processing input data according to the rule; a computation unit for reading, from the data processing unit, input data that is generated by the data processing unit, reading a model acquired by processing with the model processing unit, and acquiring intermediate data by computing, by means of this model, input data that has been processed; a model discrimination unit for computing by collating real values with verification intermediate data acquired by the processing computation, acquiring error values, and returning to a discrimination result; and a parameter adjustment unit for performing parameter adjustment on parameters of the model processing unit, the data processing unit, and the model discrimination unit.

Description

Data processing method, apparatus, device and medium

The present invention relates to the data processing technical field, and more specifically, the present invention relates to data processing methods, apparatuses, devices and media.

In recent years, models generated by machine learning have been widely applied in multiple fields. Conventional technology predicts the state of a secondary battery by learning a neural network. With the increasing use of such neural network models and the increasing number of situations in which devices on which such models reside are connected to networks, the number of situations in which third parties can connect to the models also increases.

In such cases, a third party may obtain output data by sending input data to the model. A third party can estimate the structure of the output model by acquiring many combinations of such input/output data.

To provide a data processing device that solves the technical problem that the conventional technology cannot meet the user's demand for confidentiality of the battery output model, and that the value of the model cannot be demonstrated to anyone other than authorized users. .

To achieve the above technical objectives, the present invention provides a data processing device, the data processing device comprising:
a model processing unit for reading model processing rules and performing processing on models according to the model processing rules;
a data processing unit for reading data processing rules and for processing input data according to said data processing rules;
reading processed input data generated by said data processing unit from said data processing unit; reading a model obtained by processing by said model processing unit; and using said model to compute processed input data. a computing unit for obtaining intermediate data;
a model discrimination unit for comparing intermediate data for verification obtained by processing calculation with real values, obtaining an error value, and returning to a discrimination result;
a parameter adjustment unit for performing parameter adjustments on parameters of the model processing unit, the data processing unit and the model discriminating unit.
Here, the "data processing rule" includes cases where there is no change in data.

Furthermore,
The parameter adjustment unit is
a comprehensive parameter adjustment module for generating, adjusting and optimizing each parameter of each module in the model processing unit and the data processing unit;
a parameter adjustment module of the model discrimination unit for generating, adjusting and optimizing each parameter of each module in the model discrimination unit;
a model processing rule generation module for organizing processing rules for a parameter generation model based on the model;
an input data processing rule generation module for processing a processing rule for input/output data for parameter generation based on input data.
Furthermore,
The processing rules for the parameter generation model and/or the processing rules for input/output data for parameter generation are:
Including performing transformation processing on the model and input data,
Including methods of increasing/decreasing data values, increasing/decreasing the number of values contained in input data, adjusting displacements for parameters in input data, changing values in input data, etc. independently or in combination. It is characterized by

Furthermore, the data processing unit may:
an input data read module for reading read input data;
an input data processing rule reading module for reading input data processing rules;
an input data processing module for processing the stored input data using a processing rule for the input data to obtain processed input data;
a processed input data output module for outputting processed input data;
a processed input data storage module for storing processed input data.

Further, the model processing unit may:
a model reading module for reading the model of the input model processing unit;
a model processing rule reading module for reading model processing rules;
a model processing module for processing the read model according to model processing rules to generate a processed model;
a processed model output module for outputting the processed model generated by the model processing module;
a processed model storage module for storing the processed model generated by the model processing module.

Furthermore, the computing unit may:
a post-processing model reading module that reads a post-processing model from the model processing unit;
a post-processing data reading module for reading data processed by the input data processing module;
an intermediate data calculation module for calculating the processed input data according to the model read by the processed model reading module to obtain intermediate data;
an intermediate data storage module for storing the intermediate data.

Furthermore, the model discrimination unit
an intermediate data reading module for reading intermediate data generated by the computing unit;
a model parameter reading module for reading the model parameters generated by the parameter adjustment module of the model discriminating unit;
a discriminant model calculation module for calculating the intermediate data read by the intermediate data reading module to obtain estimated data;
a discrimination result storage module for storing discrimination results of the model discrimination unit;
a discriminant result matching module for matching the estimated data with actual results and performing analytical calculations to obtain errors;
an end condition determination module that determines whether or not a determination end condition is satisfied, terminates the determination process if YES, and continues and circulates parameter adjustment if NO;
a raw output data reading module for reading output data without special processing.

Furthermore, the model discrimination unit
an intermediate data reading module for reading intermediate data generated by the computing unit;
a model parameter reading module for reading the model parameters generated by the parameter adjustment module of the model discriminating unit;
A discriminant model calculation module for matching and calculating the intermediate data read by the intermediate data reading module and the raw output data to obtain a matching result; and a discriminant result for storing the discriminant result of the model discriminating unit. a storage module;
an end condition determination module that determines whether or not a determination end condition is satisfied, terminates the determination process if YES, and continues and circulates parameter adjustment if NO;
a raw output data reading module for reading output data without special processing.

Further, errors include, but are not limited to, using standard deviation, mean squared error or 01 error as the standard of discrimination of said model discriminating unit;
The error is fed back to the model processing rule generation module, the model processing unit parameter adjustment module and the model discrimination unit parameter adjustment module.

To achieve the above technical objectives, the present invention provides a data processing method acting on the above data processing apparatus, said data processing method comprising:
using an input data processing unit to obtain input data composed of multiple values and input data processing rules generated by the parameter adjustment unit, and transform the input data based on the input data processing rules; a step of processing;
using the model processing unit to obtain a raw model and model processing rules generated by the parameter adjustment unit, and performing transformation processing on the model based on the model processing rules;
using the computing unit to compute the input data after being processed according to the input data processing rules by the model after being processed according to the model processing rules, comprising a plurality of values and having multiple rows and multiple columns; a step of outputting intermediate data composed of data;
using the model discriminating unit to read the calculated and obtained estimated data, compare and calculate with the obtained actual result, obtain the error and return to the discriminated result.

Furthermore, the data processing method includes:
using the model discriminating unit to determine whether the discriminating end condition is met, if YES, terminate the discriminating process, output the model after training, and if NO, continue to cycle through parameter adjustment; Generating and collating intermediate data.

To achieve the above technical object, the present invention provides a computer storage medium storing a computer program, which implements steps corresponding to the above data processing method when the computer program is executed by a processor. Further provided can be a computer storage medium used for.

To achieve the above technical objectives, the present invention provides an electronic device comprising a memory, a processor and a computer program stored in the memory and operable on the processor, wherein when the processor executes the computer program, the data Further provided is an electronic device in which the steps corresponding to the processing method are implemented.

1 shows a structural schematic diagram of the device of Example 1 of the present invention; FIG. FIG. 4 shows a schematic diagram of a conversion rule generated by a rule unit according to Example 1 of the present invention; FIG. 4 shows a schematic diagram of a conversion rule generated by a rule unit according to Example 1 of the present invention; FIG. 4 shows a schematic diagram of a conversion rule generated by a rule unit according to Example 1 of the present invention; FIG. 4 shows a schematic diagram of a conversion rule generated by a rule unit according to Example 1 of the present invention; FIG. 4 shows a schematic diagram of a conversion rule generated by a rule unit according to Example 1 of the present invention; Fig. 4 shows a structural schematic diagram of a data processing unit of the apparatus of Embodiment 1 of the present invention; Fig. 3 shows a structural schematic diagram of a model processing unit of the apparatus of Example 1 of the present invention; Fig. 4 shows a structural schematic diagram of the parameter adjusting unit of the apparatus of Example 1 of the present invention; Fig. 4 shows a structural schematic diagram of a computing unit of the device of Embodiment 1 of the present invention; Fig. 4 shows a structural schematic diagram of a model discriminating unit of the apparatus of Example 1 of the present invention; Fig. 4 shows a structural schematic diagram of a model discriminating unit of the apparatus of Example 1 of the present invention; 2 shows a flow chart of a second embodiment of the present invention. 2 shows a flow chart of a second embodiment of the present invention. Fig. 4 shows a structural schematic diagram of Embodiment 4 of the present invention;

Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, it should be understood that these descriptions are illustrative only and do not limit the scope of the invention. In the following description, descriptions of known structures and techniques are omitted to avoid unnecessary confusion with the concepts of the present invention.

The drawings show structural schematic diagrams according to embodiments of the present invention. These drawings are not drawn to scale, and some details may have been enlarged and some details omitted for clarity of purpose. The shapes of the various regions and layers shown in the figures and the relative sizes and positions therebetween are merely exemplary, and in practice variations may occur due to manufacturing tolerances or technical limitations. Yes, and those skilled in the art can design different regions/layers with different shapes, sizes and relative positions according to actual requirements, or according to the actual situation, to the logical relationship and/or topological structure in the drawings. may be adjusted.
Example 1

As shown in FIG. 1(a),
The present invention provides a data processing device, the data processing device comprising:
a model processing unit 101 for reading model processing rules and performing processing on models according to said model processing rules;
a data processing unit 102 for reading output data processing rules and for processing input data according to said data processing rules;
for reading input data after processing by rules from said data processing unit, reading a model obtained by training by said model processing unit, and using this model to calculate post-processing input data to obtain intermediate data; a computing unit 103;
a model discrimination unit 104 for comparing the verification intermediate data acquired by the processing calculation with the real value, acquiring the error value, and returning to the discrimination result;
a parameter adjustment unit 105 for performing parameter adjustments on the parameters of the model processing unit, the data processing unit and the model discriminating unit.
The intermediate data obtained by the data processing device shown in FIG. 1(a) is non-significant, that is, meaningless data. However, a legitimate user who understands the processing rule can obtain a correct result by performing reverse calculation (reverse operation of the processing rule) on the intermediate data. FIG. 2(b) shows this as a concept, in which a back calculation unit that performs back calculation on the output of the calculation unit is connected to obtain a correct result. However, it is only shown as a concept, and the hardware configuration of the calculation unit and the back calculation unit are not necessarily physically connected. Rather, based on what has been described in the background art and problems, for example, in the form of SaaS, a form in which a calculation unit is arranged on the server side and an inverse calculation unit is arranged on the client side is assumed.

Furthermore, as shown in FIG.
The parameter adjustment unit 105 is
a comprehensive parameter tuning module for generating, tuning and optimizing each parameter of each module in the model processing and data processing unit;
a parameter adjustment module of the model discriminating device for generating, adjusting, and optimizing each parameter of each module in the model discriminating unit;
a model processing rule generation module for organizing processing rules for a parameter generation model based on the model;
an input data processing rule generation module for processing a processing rule for input/output data for parameter generation based on input data.
Furthermore,
The processing rules for the parameter generation model and/or the processing rules for input/output data for parameter generation are:
Including performing transformation processing on the model and input data,
Including methods of increasing/decreasing data values, increasing/decreasing the number of values contained in input data, adjusting displacements for parameters in input data, changing values in input data, etc. independently or in combination. It is characterized by

As shown in FIG. 2, the computation unit uses the read processed model to determine the processed input data (to which data processing rules that do not modify the data in any way are applied), i.e. the raw input data. data to obtain intermediate data for verification, where the encryption processing rule of the model takes a specific matrix column exchange as an example, and the post-processing input data (1 2 3) is processed by the model After calculating, the intermediate data for verification (29 23 35) is obtained, and the user uses this specific matrix column exchange rule to perform an inverse operation on the intermediate data (here, the first position data of the intermediate data and second position data), the actual output data (23 29 35) cannot be obtained.
Only authorized users who understand the processing rules can obtain correct results by connecting back calculation units (not shown).

As shown in FIG. 3, the computation unit computes the processed input data (to which data processing rules that do not change the data in any way) are applied, i.e. the raw input data, according to the read processed model. to obtain intermediate data for verification, where the encryption processing rule of the model takes a specific matrix row exchange as an example, and the post-processing input data (1 2 3) is calculated by the post-processing model and then verified intermediate data (25 31 35) for .
Only authorized users who understand the processing rules can obtain correct results by connecting back calculation units (not shown).

As shown in FIG. 4, the computation unit computes the processed input data (to which data processing rules that do not change the data in any way) are applied, i.e. the raw input data, according to the read processed model. to obtain intermediate data for verification, where the encryption processing rule of the model is to exchange a certain number of positions in the specified matrix with each other as an example, and after processing the input data (1 2 3) is processed If the intermediate data for verification (23 31 35) is obtained after being calculated by the model after the calculation, and the user does not operate backward on said intermediate data using this particular matrix row exchange rule, the actual output Unable to get data (23 29 35).
Only authorized users who understand the processing rules can obtain correct results by connecting back calculation units (not shown).

As shown in FIG. 5, the computation unit computes the processed input data read by the processed model read to obtain intermediate data for verification, where the input data processing rule is a specific extension Taking the data as an example, at the same time, the model also undergoes a corresponding transformation. Intermediate data for verification (23 0 29 0 35) is obtained by calculation using the model after processing.

Unless the user uses this specific transformation rule to perform an inverse operation on the intermediate data (here, ignoring 0 values in the intermediate data and reducing the size of the vector of the intermediate data), the actual Unable to get output data (23 29 35).
Only authorized users who understand the processing rules can obtain correct results by connecting back calculation units (not shown).

As an example of the data processing method, as shown in FIG.
After the raw input data is calculated for the raw model, the raw output data is obtained;
Add two numbers to the raw input data to get the post-processing input data,
A processed model is obtained by processing the raw model with the model processing rules.
After processing the input data, after obtaining the intermediate data after calculating the post-processing model,
Inputting a later discriminant model into the raw output data and the intermediate data for verification, it is possible to determine whether the processing rules perform the encryption role well.

In addition, using this method, only if the user understands the processing rules after encryption using this method, it is possible to perform inverse arithmetic processing of the processing rules on the intermediate data, and correct results can be obtained. Even if a malicious user steals the entire model, since the processing rules are not known, only meaningless intermediate data can be obtained, and the value of the model cannot be demonstrated. It serves to protect the model and prevent malicious theft.

Further, as shown in FIG. 7, the data processing unit 102 may:
an input data read module for reading read input data;
an input data processing rule reading module for reading input data processing rules;
an input data processing module for processing the stored input data using the input data processing rules to obtain processed input data;
a processed input data output module for outputting processed input data;
a processed input data storage module for storing processed input data.

Further, as shown in FIG. 8, the model processing unit 101 may:
a model reading module for reading the model of the input model processing unit;
a model processing rule reading module for reading model processing rules;
a model processing module for processing the read model according to model processing rules to generate a processed model;
a processed model output module for storing a processed model generated by the model processing module;
a processed model storage module for outputting a processed model generated by the model processing module.

Further, as shown in FIG. 10, the computing unit 103 may:
a post-processing model reading module that reads a post-processing model from the model processing unit;
a post-processing data reading module for reading data processed by the input data processing module;
an intermediate data calculation module for calculating the processed input data according to the model read by the processed model reading module to obtain intermediate data;
an intermediate data storage module for storing the intermediate data.

Furthermore, as shown in FIG. 11, the model discrimination unit 104
an intermediate data reading module for reading intermediate data generated by the computing unit;
a model parameter reading module for reading the model parameters generated by the parameter adjustment module of the model discriminating unit;
a discriminant model calculation module for calculating the intermediate data read by the intermediate data reading module to obtain estimated data;
a discrimination result storage module for storing discrimination results of the model discrimination unit;
a discriminant result matching module for matching the estimated data with actual results and performing analytical calculations to obtain errors;
an end condition determination module that determines whether or not a determination end condition is satisfied, terminates the determination process if YES, and continues and circulates parameter adjustment if NO;
a raw output data reading module for reading output data without special processing.

Further, as shown in FIG. 12, in another embodiment, the model discriminating unit 104:
an intermediate data reading module for reading intermediate data generated by the computing unit;
a model parameter reading module for reading the model parameters generated by the parameter adjustment module of the model discriminating unit;
A discriminant model calculation module for matching and calculating the intermediate data read by the intermediate data reading module and the raw output data to obtain a matching result; and a discriminant result for storing the discriminant result of the model discriminating unit. a storage module;
an end condition determination module that determines whether or not a determination end condition is satisfied, terminates the determination process if YES, and continues and circulates parameter adjustment if NO;
a raw output data reading module for reading output data without special processing.

Further, said error includes, but is not limited to, using standard deviation, mean squared error or 01 error as the standard of discrimination of the model discriminating unit;
The error is fed back to the model processing rule generation module, the model processing unit parameter adjustment module and the model discrimination unit parameter adjustment module.
Example 2

As shown in FIG.
A data processing method acts on the above data processing apparatus, said data processing method comprising: a.
S201: using the data processing unit to acquire input data composed of a plurality of values and processing rules generated by the model processing unit, and perform transformation processing on the input data based on the processing rules; step to do.
S202: Outputting intermediate data composed of multiple values and composed of multiple rows or multiple columns of data by performing processing on the processed input data using the computing unit.
S203: Using the model discriminating unit to read the calculated and obtained estimated data, compare and calculate with the obtained actual result, obtain the error and return to the discriminated result.

Furthermore, the data processing method includes:
Use the model discriminating unit to determine whether the discriminating end condition is met, if YES, terminate the discriminating process, output the model after training, and if NO, continue parameter adjustment. Including cycling steps.

A data processing method acting on a parameter adjustment unit, said parameter adjustment method comprising: a.
Reading the calculation results by the comprehensive parameter tuning module and adjusting the parameters of each model module based on the feedback of the calculation results.
A step of generating, adjusting, and optimizing each parameter of each module in the model processor by the parameter adjustment module of the model processor.
A step of generating, adjusting, and optimizing each parameter of each module in the model discriminating device by the parameter adjustment module of the model discriminating device.
generating, adjusting, and optimizing rules for model processing, and generating input/output data processing rules by a model processing rule generation module; Here, the processing means conversion processing for the model, and the conversion processing method is an increase/decrease in the quantity of values included in the input data due to an increase/decrease in the data, or an increase in the quantity of values included in the input data. This includes, but is not limited to, methods that use /decrease, displacement adjustments to parameters in the input data, changes in values in the input data, etc. independently or in combination.

The data processing method acts on the model processing unit, the model processing method obtains an input model and model processing rules, adjusts and trains the model according to the model processing rules, and produces a model after adjustment processing. Including the output step.
in particular,
reading the raw model with a model reading module;
reading model processing rules with a model processing rules reading module;
processing, by a model processing module, the read model in the manner of model processing rules to generate a new processed model.
A model storage module stores the processed model generated by the model processing model.
Read the raw input data by the input data reading module,
read the input data processing rule by the input data processing rule reading module;
The input data processing module processes the read input data according to the input data processing rules to generate new processed input data.
An input data storage module stores the processed input data generated by the input data processing model.
The intermediate data calculation module calculates the processed input data stored in the processed input data storage module using the processed model stored in the processed model storage module to obtain intermediate data.

A data processing method acts on the computing unit, the computing method comprising:
read the input data processed according to the input data processing rules from the input data processing unit, read the model processed according to the model processing rules from the model processing unit, and use the model to process the processed input data. Including calculating and obtaining intermediate data. Specifically, it includes:
The module is read by the processed model, and the processed model is read from the model processor.
The module is read with post-processing data, and the input data is read after being processed by the data processor described below.
An intermediate data calculation module calculates the processed input data using the model read by the model reading module to obtain intermediate data.
The intermediate data storage module stores intermediate data generated after calculation by the model calculation module to which it belongs.

A data processing method is used in a model discrimination unit, said model discrimination method comprising:
Using the model discriminant model to calculate the intermediate data calculated and obtained by the model processing device, obtain estimated output data, compare with the actual output data and calculate, and obtain an error value. including. in particular,
An intermediate data reading module reads the intermediate data generated by the model processor.
The model parameter acquisition module reads the model parameters generated by the parameter adjustment module of the model discriminator.
A model storage module stores the model of the model discriminator.
The model calculation module calculates the intermediate data acquired by the intermediate data reading module using the model of the model discriminating device, and acquires an estimation result.
A result matching module matches the estimated result with the actual result and calculates to obtain the error. Specifically, errors herein include, but are not limited to, using standard deviations, mean squared errors, 01 errors, and the like. The error is fed back to the model processing rule generation module, the parameter adjustment module of the model processor, and the parameter adjustment module of the model discriminator.
The termination condition determination module determines whether or not the entire flow has achieved the termination condition. Termination conditions include, but are not limited to, reaching a certain number of training times and having a distribution of the entire encrypted data within a certain range.
The artificial identification error rate exceeds a certain threshold. If the termination condition is satisfied, the flow is terminated, and if not satisfied, the next flow is circulated.

As shown in FIG.
First, based on preset comprehensive parameters (including model processing rules and input data processing rules), input data processing rules and model processing rules are generated, respectively, and then input data processing and model processing are performed, respectively, to obtain input data. Calculation is performed using the post-processed input data and the post-processed model, and intermediate data for verification is acquired. Intermediate data for verification is calculated by a discrimination model to obtain a discrimination result, and it is determined whether or not a process termination condition is satisfied based on the discrimination result and conditions such as parameters in the calculation. If so, the operation of the device is terminated after saving each parameter. If not, after adjusting the parameters of the model discriminator, perform the comprehensive parameter adjustment again and start the cycle.
Example 3

The present invention can further provide a computer storage medium storing a computer program, which is used to implement the steps of the above data processing method when the computer program is executed by a processor.

The computer storage medium of the present invention can be implemented using semiconductor memory, magnetic core memory, magnetic drum memory, or magnetic disk memory.

Semiconductor memories are mainly used as semiconductor memory elements in computers, and there are mainly two types: MOS and bipolar types. MOS devices have a high degree of integration and a simple process, but are slow. Bipolar devices are more complicated to process, consume more power, have lower integration, but are faster. After the advent of NMOS and CMOS, MOS memory began to occupy a dominant position in semiconductor memory. NMOS is fast, for example, Intel's 1 Kbit static random memory has an access time of 45 ns. CMOS consumes less power and the access time of 4Kbit CMOS static memory is 300ns. All of the above semiconductor memories are random access memories (RAMs), ie, new contents can be read and written randomly in the process of operation. Semiconductor read-only memory (ROM) can be randomly read in the process, but cannot be written to, and is used to store fixed programs and data.

ROM is also divided into two types: non-rewritable fuse-type read-only memory -- PROM and rewritable read-only memory EPROM.

Magnetic core memory has low cost, high reliability, and more than 20 years of practical experience. Until the mid-70's, magnetic core memory was widely used as main memory. Its storage capacity can reach more than 10 bits and the fastest access time is 300ns. The typical magnetic core memory capacity in the world is 4MS~8MB, and the access cycle is 1.0~1.5μs. After the rapid development of semiconductor memory to replace the magnetic core memory as the main memory, the magnetic core memory can still be applied as a large-capacity expansion memory.

A magnetic drum memory is an external memory for magnetic recording. Because of its fast information access speed, stable and reliable operation, its capacity is relatively small, and it is gradually being replaced by magnetic disk memory. Used as external memory. In order to meet the demands of small and mini-computers, an ultra-miniature magnetic drum with small volume, light weight, high reliability and convenient use has emerged.

A magnetic disk memory is an external memory for magnetic recording. It combines the advantages of magnetic drums and magnetic tape memories, that is, its storage capacity is larger than that of magnetic drums, its access speed is faster than that of magnetic tape memories, and it can be stored offline. It is widely used as a large-capacity external memory. Magnetic disks are generally divided into two types: hard disks and floppy disk memories.

There are many types of hard disk memory. Structurally, there are two types: removable type and fixed type. The platters of removable magnetic disks are replaceable and the platters of fixed magnetic disks are fixed. Both removable and fixed magnetic disks have two types, multi-piece combination and single-piece construction, and are divided into fixed magnetic head type and movable magnetic head type respectively. A fixed magnetic head type magnetic disk has a small capacity, a low recording density, and a high access speed, but the production cost is high. A movable magnetic head type magnetic disk has a high recording density (1000 to 6250 bits/inch) and thus a large capacity, but its access speed is lower than that of a fixed magnetic head type magnetic disk. The storage capacity of magnetic disk products can reach hundreds of megabytes, with a bit density of 6250 bits/inch and a track density of 475 tracks/inch. Here, the multi-piece removable magnetic disk memory has a combination of disks that can be exchanged, has a large offline capacity, has a large capacity, a high speed, and can store a large amount of information materials, so it can be used for online information retrieval. Widely applied to systems, database management systems.
Example 4

The present invention is an electronic device comprising a memory, a processor and a computer program stored in the memory and operable on the processor, wherein the steps of the above data processing method are realized when the processor executes the computer program. further provide.

FIG. 15 is a schematic diagram of the internal structure of an electronic device in one example. As shown in FIG. 15, the electronic device includes a processor, storage medium, memory and network interface connected via a system bus. Here, the storage medium of the computing device stores an operating system, a database and computer readable instructions, the database capable of storing widget information sequences, and the processor executing the computer readable instructions. , the processor can implement the data processing method. The electronic device's processor is used to provide computing and control capabilities and to support the operation of the entire computing device. The memory of the computing device may store computer readable instructions which, when executed by the processor, may cause the processor to perform data processing methods. A network interface of this computing device is used for connection communication with a terminal. The structure shown in FIG. 15 is merely a block diagram of some structures related to aspects of the present application, and does not limit computer devices to which aspects of the present application may be applied. One skilled in the art will appreciate that it may include more or fewer members than those shown, or may be combined with certain members or have a different arrangement of members.

This electronic device includes, but is not limited to, smartphones, computers, tablet computers, wearable smart devices, artificial intelligence devices, power banks, etc.

In an embodiment, the processor may be composed of an integrated circuit, for example, may be composed of a single package integrated circuit, may be composed of a plurality of same or different function package integrated circuits, It includes one or more Central Processing Units (CPUs), microprocessors, digital processing chips, graphics processors and combinations of various control chips. The processor is a control unit of the electronic device, uses various interfaces and circuits to connect each member of the entire electronic device, and programs or modules stored in the memory (for example, It performs various functions of the electronic device and processes data by running or executing programs, such as programs that read and write remote data, and retrieving data stored in the memory.

The bus may be a peripheral component interconnect (abbreviated as PCI) bus or an extended industry standard architecture (abbreviated as EISA) bus, or the like. This bus is divided into an address bus, a data bus, a control bus, and the like. The bus is provided for providing connection communication between the memory and at least one processor or the like.

FIG. 15 shows only an electronic device having members, and the structure shown in FIG. 15 is not intended to limit said electronic device and may include fewer or more members than those shown in the drawing, or may be combined with certain members, One skilled in the art will appreciate that it may have different member arrangements.

For example, although not shown, the electronic device may further include a power source (e.g., battery) for powering each component, preferably the power source is logically connected to the at least one processor via a power management unit. As a result, the power management device implements functions such as charge management, discharge management, and power consumption management. The power source may further include optional components such as one or more DC or AC power sources, rechargers, power failure detection circuitry, power converters or inverters, power status indicators, and the like. The electronic device may include multiple sensors, a Bluetooth module, a Wi-Fi module, etc., and will not be described here.

Further, the electronic device may further include a network interface, optionally the network interface may include a wired interface and/or a wireless interface (eg, WI-FI interface, Bluetooth interface, etc.), typically It is used to establish communication connections between this electronic device and other electronic devices.

Optionally, the electronic device may further include a user interface, the user interface may be a display, an input unit (e.g., a keyboard), optionally the user interface may include a standard It may be a wired interface or a wireless interface. Alternatively, in some embodiments, the display may be an LED display, a liquid crystal display, a touch liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, and the like. Here, the display, which may be appropriately referred to as a display panel or display unit, is used to display processed information in the electronic device and is used to display a visualized user interface.

Further, the computer-usable storage medium can mainly include a storage program zone and a storage data zone, where the storage program zone contains an operating system, application programs required for at least one function, etc. The storage data zone can store data, etc. created based on the blockchain node.

It should be understood that in some of the embodiments provided by the present invention, the disclosed devices, apparatus and methods may be embodied in other forms. For example, the apparatus embodiments described above are exemplary only, e.g., the partitions of the modules are only partitions of logical functions, and other partition forms are possible when actually implemented. .

The modules described above as separate members may or may not be physically separated, and the members indicated as modules are physical units. It may or may not be a physical unit, ie it may be located in one place or distributed over several network units. Based on actual needs, some or all of the modules may be selected to achieve the purpose of this aspect of the embodiment.

Each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically independently, and two or more units may be integrated into one processing unit. may be integrated. The above integrated units may be implemented in the form of hardware or in the form of functional modules combining hardware and software.

The embodiments of the present invention have been described above. However, these examples are for illustrative purposes only and are not intended to limit the scope of the invention. The scope of the invention is defined by the Claims appended hereto and their equivalents. Various replacements and modifications can be made by those skilled in the art without departing from the scope of the present invention, and all of these replacements and modifications should fall within the scope of the present invention.

101 model processing unit 102 data processing unit 103 calculation unit 104 model discrimination unit 105 parameter adjustment unit

Claims

a model processing unit for reading model processing rules and performing processing on models according to the model processing rules;
a data processing unit for reading data processing rules and for processing input data according to said data processing rules;
reading processed input data generated by said data processing unit from said data processing unit; reading a model obtained by processing by said model processing unit; and using said model to compute processed input data. a computing unit for obtaining intermediate data through
a model discrimination unit for comparing intermediate data for verification obtained by processing calculation with real values, obtaining an error value, and returning to a discrimination result;
a parameter adjustment unit for adjusting parameters of the model processing unit, the data processing unit and the model discriminating unit.
The parameter adjustment unit is
a comprehensive parameter adjustment module for generating, adjusting and optimizing each parameter of each module in the model processing unit and the data processing unit;
a parameter adjustment module of the model discrimination unit for generating, adjusting and optimizing each parameter of each module in the model discrimination unit;
a model processing rule generation module for organizing processing rules for a parameter generation model based on the model;
2. The apparatus of claim 1, comprising an input data processing rule generating module for processing input data processing rules for generating parameters based on the input data.
The processing rules for the parameter generation model and/or the processing rules for input/output data for parameter generation are:
Including performing transformation processing on the model and input data,
Including methods of increasing/decreasing data values, increasing/decreasing the number of values contained in input data, adjusting displacements for parameters in input data, changing values in input data, etc. independently or in combination. 3. Apparatus according to claim 2, characterized in that:
The data processing unit is
an input data read module for reading read input data;
an input data processing rule reading module for reading input data processing rules;
an input data processing module for processing the stored input data using a processing rule for the input data to obtain processed input data;
a processed input data output module for outputting processed input data;
4. Apparatus according to claim 2 or 3, comprising a processed input data storage module for storing processed input data.
The model processing unit includes:
a model reading module for reading the model of the input model processing unit;
a model processing rule reading module for reading model processing rules;
a model processing module for processing the read model according to model processing rules to generate a processed model;
a processed model output module for outputting the processed model generated by the model processing module;
5. The apparatus of claim 4, comprising a processed model storage module for storing the processed model generated by the model processing module.
The computing unit is
a post-processing model reading module that reads a post-processing model from the model processing unit;
a post-processing data reading module for reading data processed by the input data processing module;
an intermediate data calculation module for calculating the processed input data according to the model read by the processed model reading module to obtain intermediate data;
and an intermediate data storage module for storing the intermediate data.
The model discrimination unit is
an intermediate data reading module for reading intermediate data generated by the computing unit;
a model parameter reading module for reading the model parameters generated by the parameter adjustment module of the model discriminating unit;
a discriminant model calculation module for calculating the intermediate data read by the intermediate data reading module to obtain estimated data;
a discrimination result storage module for storing discrimination results of the model discrimination unit;
a discriminant result matching module for matching the estimated data with actual results and performing analytical calculations to obtain errors;
an end condition determination module that determines whether or not a determination end condition is satisfied, terminates the determination process if YES, and continues and circulates parameter adjustment if NO;
7. The apparatus of claim 6, comprising a raw output data reading module for reading output data that has not undergone special processing.
The model discrimination unit is
an intermediate data reading module for reading intermediate data generated by the computing unit;
a model parameter reading module for reading the model parameters generated by the parameter adjustment module of the model discriminating unit;
A discriminant model calculation module for matching and calculating the intermediate data read by the intermediate data reading module and the raw output data to obtain a matching result; and a discriminant result for storing the discriminant result of the model discriminating unit. a storage module;
an end condition determination module that determines whether or not a determination end condition is satisfied, terminates the determination process if YES, and continues and circulates parameter adjustment if NO;
7. The apparatus of claim 6, comprising a raw output data reading module for reading output data that has not undergone special processing.
The error includes, but is not limited to, using standard deviation, mean squared error or 01 error as the standard of discrimination of the model discriminating unit;
8. The apparatus of claim 7, wherein the error is fed back to the model processing rule generation module, the model processing unit parameter adjustment module and the model discrimination unit parameter adjustment module.
A data processing method acting on a data processing apparatus according to any one of claims 1 to 9, comprising:
using an input data processing unit to obtain input data composed of multiple values and input data processing rules generated by the parameter adjustment unit, and transform the input data based on the input data processing rules; a step of processing;
using the model processing unit to obtain a raw model and model processing rules generated by the parameter adjustment unit, and performing transformation processing on the model based on the model processing rules;
using the computing unit to compute the input data after being processed according to the input data processing rules by the model after being processed according to the model processing rules, comprising a plurality of values and having multiple rows and multiple columns; a step of outputting intermediate data composed of data;
using the model discriminating unit to read the calculated and obtained estimated data, compare and calculate with the obtained actual result, obtain the error and return to the discriminated result. data processing method.
The data processing method includes:
using the model discriminating unit to determine whether the discriminating end condition is met, if YES, terminate the discriminating process, output the model after training, and if NO, continue to cycle through parameter adjustment; 11. The method of claim 10, further comprising generating and collating intermediate data.
An electronic device comprising a memory, a processor and a computer program stored in the memory and operable on the processor, wherein the steps corresponding to the data processing method according to claim 10 or 11 are realized when the processor executes the computer program. electronic device.
A computer storage medium storing computer program instructions, characterized in that, when executed by a processor, the computer implements the steps corresponding to the data processing method of claim 10 or 11. storage medium.