Incremental analysis-based causal reasoning method and device, equipment, medium and product
By introducing incremental analysis technology into Bayesian causal reasoning method, the control causal graph is constructed and the probability distribution of the target causal path is updated, the problem of inefficient causal reasoning is solved and efficient causal reasoning is achieved.
Patent Information
- Application Number
- CN202510043907.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-10
- Publication Date
- 2025-05-16
AI Technical Summary
As the variable size becomes larger, the reasoning efficiency of Bayesian causal reasoning method is restricted and it is impossible to effectively deal with the causal reasoning task of a large number of variables.
The causal reasoning method based on incremental analysis is adopted, and the probability distribution of the affected target causal path is obtained by constructing a control causal graph, and the incremental analysis intervention variable is determined based on the control causal graph and intervention variable, and only the probability distribution of the affected target causal path is updated.
It improves the efficiency of causal reasoning, can effectively handle causal reasoning tasks of a large number of variables, and reduces the calculation amount and reasoning time.
Smart Images
Figure CN120012921A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a causal reasoning method, device, equipment, medium and product based on incremental analysis. Background Art
[0002] In causal models, the common problem of inference about frequent interventions is particularly important.
[0003] At present, in the existing technology, the causal reasoning method is usually based on Bayesian reasoning. The Bayesian reasoning method can perform causal reasoning in a static or dynamic manner, but as the scale of variables increases, the Bayesian causal reasoning method needs to re-reason each time, which restricts the reasoning efficiency and reduces the reasoning efficiency.
[0004] Therefore, there is an urgent need for an incremental analysis causal reasoning method to improve reasoning efficiency. Summary of the invention
[0005] The present invention provides a causal reasoning method, device, equipment, medium and product based on incremental analysis, which is used to solve the defect in the prior art that the reasoning efficiency of the Bayesian causal reasoning method is restricted as the scale of variables increases. It can determine the affected target causal path based on candidate causal paths and incremental analysis intervention variables, and only update the probability distribution corresponding to the affected target causal path, thereby improving the reasoning efficiency.
[0006] The present invention provides a causal reasoning method based on incremental analysis, comprising the following steps.
[0007] A control causal graph is constructed; wherein the control causal graph includes all probability distributions corresponding to all causal variables and all candidate causal paths determined by all causal variables.
[0008] Obtain incremental analysis intervention variables; wherein incremental analysis intervention variables refer to intervention variables that affect candidate causal paths.
[0009] All target causal paths are determined based on the control causal diagram and the incremental analysis intervention variables; wherein all target causal paths are the paths affected by the incremental analysis intervention variables among all candidate causal paths in the control causal diagram.
[0010] The probability distributions corresponding to the target causal paths are updated respectively to determine the joint distribution results of causal reasoning.
[0011] According to a causal reasoning method based on incremental analysis provided by the present invention, a control causal graph is constructed, including: obtaining a causal graph; wherein the causal graph contains all causal variables and the dependencies between all causal variables; and constructing a control causal graph according to the causal variables and the dependencies in the causal graph.
[0012] According to a causal reasoning method based on incremental analysis provided by the present invention, a control causal graph is constructed according to causal variables and dependency relationships in the causal graph, including: determining an initial causal path according to the causal variables in the causal graph; determining a candidate causal path according to the dependency relationship between the initial causal path and all causal variables; obtaining all initial probability distributions corresponding to all causal variables; and constructing a control causal graph according to the candidate causal paths and all initial probability distributions.
[0013] According to a causal reasoning method based on incremental analysis provided by the present invention, all target causal paths are determined according to a control causal graph and incremental analysis intervention variables, including: determining whether the intervention variable branch corresponding to the intervention variable in the incremental analysis intervention variable satisfies the causal variable branch corresponding to the causal variable in all candidate causal paths in the control causal graph; in the case where there is an intervention variable branch that does not satisfy the causal variable branch in all candidate causal paths, determining the candidate causal path corresponding to the causal variable branch that does not satisfy the target causal path; and determining all target causal paths according to the candidate causal paths corresponding to the causal variable branches that do not satisfy the causal variable branches of all intervention variable branches.
[0014] According to a causal reasoning method based on incremental analysis provided by the present invention, the method also includes: when there is an intervention variable branch that satisfies the causal variable branch in all candidate causal paths, the candidate causal path corresponding to the causal variable branch that satisfies the causal variable branch is determined as the remaining causal path; wherein the remaining causal path is the path among all candidate causal paths except all target causal paths.
[0015] According to a causal reasoning method based on incremental analysis provided by the present invention, each probability distribution corresponding to each target causal path is updated respectively to determine the joint distribution result of causal reasoning, including: obtaining the joint probability distribution corresponding to the remaining causal paths; wherein the joint probability distribution corresponding to the remaining causal paths remains unchanged; probability updating each probability distribution corresponding to each target causal path respectively to determine the target probability distribution corresponding to each target causal path; and determining the joint distribution result of causal reasoning according to the sum of all target probability distributions and the joint probability distribution.
[0016] The present invention also provides a causal reasoning device based on incremental analysis, comprising the following modules.
[0017] A construction module is used to construct a control causal graph; wherein the control causal graph includes all probability distributions corresponding to all causal variables and all causal paths determined by all causal variables.
[0018] The variable acquisition module is used to obtain the incremental analysis intervention variables; wherein the incremental analysis intervention variables refer to the intervention variables that affect the candidate causal path.
[0019] The path determination module is used to determine all target causal paths according to the control causal diagram and the incremental analysis intervention variables; wherein all target causal paths are the paths affected by the incremental analysis intervention variables among all candidate causal paths in the control causal diagram.
[0020] The result determination module is used to update the probability distributions corresponding to the target causal paths respectively and determine the joint distribution results of causal reasoning.
[0021] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the processor implements any of the above-mentioned causal reasoning methods based on incremental analysis.
[0022] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned causal reasoning methods based on incremental analysis.
[0023] The present invention also provides a computer program product, including a computer program, which implements any of the above-mentioned causal reasoning methods based on incremental analysis when executed by a processor.
[0024] The present invention provides a causal reasoning method, device, equipment, medium and product based on incremental analysis, which constructs a control causal graph; wherein the control causal graph contains all probability distributions corresponding to all causal variables and all candidate causal paths determined by all causal variables; obtains incremental analysis intervention variables; wherein the incremental analysis intervention variables refer to intervention variables that affect the candidate causal paths; determines all target causal paths according to the control causal graph and the incremental analysis intervention variables; wherein all target causal paths are paths affected by the incremental analysis intervention variables in all candidate causal paths in the control causal graph; updates each probability distribution corresponding to each target causal path, and determines the joint distribution result of causal reasoning. The technical solution of the present invention is used to solve the defect that the reasoning efficiency of the Bayesian causal reasoning method is restricted as the scale of variables increases in the prior art, and determines the affected target causal paths based on the candidate causal paths and the incremental analysis intervention variables, and only updates the probability distribution corresponding to the affected target causal paths, thereby improving the reasoning efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0026] Figure 1 It is one of the flow charts of the causal reasoning method based on incremental analysis provided by the present invention.
[0027] Figure 2 It is a schematic diagram of the cause-effect diagram provided by the present invention.
[0028] Figure 3 It is a schematic diagram of the control cause-effect diagram provided by the present invention.
[0029] Figure 4 This is the second flowchart of the causal reasoning method based on incremental analysis provided by the present invention.
[0030] Figure 5 It is a structural schematic diagram of a causal reasoning device based on incremental analysis provided by the present invention.
[0031] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0032] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0033] Combine the following Figure 1-Figure 4 The causal reasoning method based on incremental analysis provided by the present invention is described. The causal reasoning method based on incremental analysis provided by the present invention can be applicable to causal reasoning situations of incremental analysis. The executor of the method can be an electronic device, or a causal reasoning device based on incremental analysis arranged in the electronic device. The causal reasoning device based on incremental analysis can be implemented by software, hardware, or a combination of both. Figure 1 This is one of the flow charts of the causal reasoning method based on incremental analysis provided by the present invention, such as Figure 1 As shown, the method includes the following steps 101, 102, 103 and 104.
[0034] Step 101: Construct a control cause-effect diagram.
[0035] In this step, the control causal graph includes all probability distributions corresponding to all causal variables and all candidate causal paths determined by all causal variables.
[0036] In a specific implementation, constructing a control causal graph includes: obtaining a causal graph; wherein the causal graph includes all causal variables and dependencies between all causal variables; and constructing the control causal graph based on the causal variables and dependencies in the causal graph.
[0037] In a specific embodiment, a control causal graph is constructed based on causal variables and dependency relationships in the causal graph, including: determining an initial causal path based on the causal variables in the causal graph; determining a candidate causal path based on the dependency relationships between the initial causal path and all causal variables; obtaining all initial probability distributions corresponding to all causal variables; and constructing a control causal graph based on the candidate causal paths and all initial probability distributions.
[0038] In this step, Figure 2 is a schematic diagram of the cause-effect diagram provided by the present invention, such as Figure 2 As shown, the causal graph contains multiple variables, such as variable A, variable B, variable C and variable D. The lines between the variables in the causal graph represent the dependency relationship between the variables. The dependency relationship can be divided into conditional dependency and data dependency. For example, conditional dependency can be that when variable A meets certain conditions, variable B meets a certain probability distribution. , separated by conditional nodes; data dependencies can be variables, for example ,General causal diagrams present the dependency relationship between variables without distinguishing them. The role of causal diagrams is to provide a basis for subsequent incremental analysis and to build a control causal diagram. In the causal diagram, the initial causal paths are variable A-variable B-variable C-variable D and variable A-variable C-variable D.
[0039] Specifically, after obtaining the causal graph, all initial causal paths in the causal graph are determined based on all causal variables in the causal graph, and then the initial causal paths are divided into candidate causal paths based on the dependency relationship between the initial causal paths and all causal variables, and all initial probability distributions corresponding to all causal variables are obtained; the judgment conditions are determined based on the initial probability distributions, and a control causal graph is constructed based on the candidate causal paths and the judgment conditions.
[0040] In this step, the judgment condition can be, for example, a binary operator such as etc., decomposing the initial causal path into independent candidate causal paths, such as , where n represents the number of candidate causal paths after decomposition, and only data dependencies exist in the candidate causal paths after decomposition.
[0041] For example, Figure 3 is a schematic diagram of the control cause-effect diagram provided by the present invention, such as Figure 3As shown, in the control causal graph, for example, variable A represents the execution of two actions, one to the left and one to the right, then the judgment condition IF_A is determined to execute the first action, continue to execute variable B1 to the left, execute the second action, and continue to execute variable B2 to the right, then variables B1 and B2 continue to execute C (A, B) at the same time, and then continue to determine the judgment condition IF_C of variable C, and determine to continue to execute variable D1 or variable D2 according to the judgment condition IF_C, thereby forming a control causal graph. The candidate causal paths contained in the control causal graph can be, for example, (A, B1, C, D1), (A, B1, C, D2), (A, B2, C, D1), (A, B2, C, D2), this embodiment does not limit this.
[0042] In one embodiment, different candidate causal paths (A, B1, C, D1), (A, B1, C, D2), (A, B2, C, D1), The probability distribution calculation method corresponding to (A, B2, C, D2) can be, for example, , where FA_L is the logical judgment formula of the left branch of the judgment condition IF_A, for example, A<10, It is the logical judgment formula of the left branch of the judgment condition IF_C, which is not limited in this embodiment.
[0043] The advantage of this setting is that, based on the representation of the causal graph, the control flow technology in program analysis is used to model the dependencies of the variables in the causal graph, and the paths on the causal graph are separated according to the control flow nodes to make the paths independent of each other, thereby constructing a control causal graph. The paths in the control causal graph only contain the data dependencies of the variables, which effectively reduces the coupling relationship between the paths, so that the variables on the path are only related to the variables of the current path, which is convenient for subsequent intervention variable correlation analysis and rapid update of distribution.
[0044] Step 102: Obtain incremental analysis intervention variables.
[0045] In this step, the incremental analysis intervention variable refers to an intervention variable that affects the candidate causal path. The incremental analysis intervention variable may be, for example, do (A=3), which is not limited in this embodiment.
[0046] Specifically, after constructing the control causal diagram, the incremental analysis intervention variables are obtained, and based on the incremental analysis intervention variables, it is further determined whether all candidate causal paths in the control causal diagram are affected.
[0047] Step 103: Determine all target causal paths based on the control causal diagram and the incremental analysis intervention variables.
[0048] In this step, all target causal paths are paths affected by the incremental analysis intervention variables among all candidate causal paths in the control causal graph.
[0049] Specifically, after constructing the control causal diagram and obtaining the incremental analysis intervention variables, it is further determined whether all candidate causal paths in the control causal diagram are affected based on the incremental analysis intervention variables, the candidate causal paths affected by the incremental analysis intervention variables are determined as target causal paths, and all target causal paths are determined.
[0050] For example, the incremental analysis intervention variable do(x) is used to analyze the variable x that is intervened by the incremental analysis intervention variable, and the path containing the variable is analyzed, such as , is the affected target causal path. The remaining causal paths are remove , remove The paths are unaffected.
[0051] In a specific embodiment, all target causal paths are determined based on a control causal graph and incremental analysis intervention variables, including: determining whether an intervention variable branch corresponding to an intervention variable in the incremental analysis intervention variable satisfies a causal variable branch corresponding to a causal variable in all candidate causal paths in the control causal graph; in the case where there is an intervention variable branch that does not satisfy a causal variable branch in all candidate causal paths, determining the candidate causal path corresponding to the causal variable branch that does not satisfy the causal variable branch as the target causal path; and determining all target causal paths based on the candidate causal paths corresponding to the causal variable branches that do not satisfy the causal variable branches of all intervention variable branches.
[0052] In a specific implementation, it also includes: when there is an intervention variable branch that satisfies the causal variable branch in all candidate causal paths, the candidate causal path corresponding to the causal variable branch that satisfies the causal variable branch is determined as the remaining causal path; wherein the remaining causal path is the path among all candidate causal paths except all target causal paths.
[0053] Specifically, determine whether the intervention variable branch corresponding to the intervention variable in the incremental analysis intervention variable satisfies the causal variable branch corresponding to the causal variable in all candidate causal paths in the control causal graph; when there is an intervention variable branch that does not satisfy the causal variable branch in all candidate causal paths, the candidate causal path corresponding to the causal variable branch that does not satisfy the target causal path; determine all target causal paths based on the candidate causal paths corresponding to the causal variable branches that do not satisfy all intervention variable branches. When there is an intervention variable branch that satisfies the causal variable branch in all candidate causal paths, the candidate causal path corresponding to the causal variable branch that satisfies the target causal path. In the case that there is an intervention variable branch that satisfies the causal variable branch in all candidate causal paths, the candidate causal path that satisfies the target causal path is determined as the remaining causal path; wherein the remaining causal path is the path in all candidate causal paths except all target causal paths.
[0054] Exemplarily, all candidate causal paths include , , , , when the incremental analysis intervention variable is do (A=3), where , Since it is a branch with A>10, its distribution probability can be directly set to 0, because the affected path is a candidate causal path and Therefore, all target causal paths are determined as and .in, , If the intervention variable branch corresponding to the intervention variable in the incremental analysis intervention variable is satisfied, then determine , These are all the remaining causal paths, which are not limited in this embodiment.
[0055] Step 104: Update the probability distributions corresponding to the target causal paths respectively to determine the joint distribution result of causal reasoning.
[0056] Specifically, for paths without incremental analysis intervention variables, the joint probability distribution on the path remains unchanged; for paths with incremental analysis intervention variables , based on the previously stored , the updated distribution is calculated by the following formula = / .in, is the target causal path The data depends on the variables x of the incremental analysis intervention variable, which can be obtained by data dependence analysis. Finally, the causal inference joint distribution results . represents the sequence number of the target causal path, and n represents the total number of candidate causal paths.
[0057] In a specific implementation, each probability distribution corresponding to each target causal path is updated respectively to determine the joint distribution result of causal reasoning, including: obtaining the joint probability distribution corresponding to the remaining causal paths; wherein the joint probability distribution corresponding to the remaining causal paths remains unchanged; probability updating each probability distribution corresponding to each target causal path respectively to determine the target probability distribution corresponding to each target causal path; and determining the joint distribution result of causal reasoning based on the sum of all target probability distributions and the joint probability distribution.
[0058] Specifically, after determining the target causal path, the target probability distribution corresponding to each target causal path is determined based on the joint probability distribution of the previously stored target causal paths. The joint distribution result of causal reasoning is determined based on the sum of all target probability distributions and the joint probability distribution corresponding to the remaining causal paths.
[0059] For example, the target causal path previously stored may be The joint probability distribution of , target causal path The joint probability distribution of , target causal path The corresponding target probability distribution is ,in, In the case of incremental analysis intervention variable do (A=3), variable C passes through variables A and variable The probability distribution of represents the probability distribution of variable A<3, Indicates that variable C passes through variables A and variables The probability distribution of the target causal path The target probability distribution is Among them, the remaining causal path and The joint probability distribution of remains unchanged, that is, the remaining causal paths The target probability distribution is equal to the remaining causal paths The joint distribution probability = , the remaining causal paths The target probability distribution is equal to the remaining causal paths The joint distribution probability = Therefore, the joint distribution result of causal reasoning is equal to the sum of the joint probability distributions of all target probability distributions and the joint probability distributions corresponding to the remaining causal paths to determine the joint distribution result of causal reasoning, that is, the joint distribution result of causal reasoning , Indicates the sequence number of the target causal path.
[0060] In a specific embodiment, Figure 4 This is the second flow chart of the causal reasoning method based on incremental analysis provided by the present invention, such as Figure 4 As shown, the specific steps include the following steps.
[0061] Step 401: Control cause-effect graph construction.
[0062] Specifically, a control causal graph is constructed based on the causal graph.
[0063] Step 402: Path identification.
[0064] Specifically, after constructing the control causal diagram, path identification is performed based on the control causal diagram to determine all causal variables and judgment conditions.
[0065] Step 403: Calculate path distribution.
[0066] Specifically, path distribution calculation is performed through all causal variables and judgment conditions to determine all candidate causal paths.
[0067] Step 404: Intervention.
[0068] Specifically, after all candidate causal paths are determined, the candidate causal paths are further intervened to obtain incremental analysis intervention variables, and the candidate causal paths are intervened based on the incremental analysis intervention variables.
[0069] Step 405: Analyze the affected paths.
[0070] Specifically, the candidate causal paths are intervened based on the intervention variables of the incremental analysis, and the affected path analysis is performed to determine the affected candidate causal paths.
[0071] Step 406: Update the path distribution.
[0072] Specifically, the path distribution is updated based on the candidate causal paths affected by the incremental analysis intervention variables and the candidate causal paths not affected by the incremental analysis intervention variables, so that the candidate causal paths affected by the incremental analysis intervention variables are determined as the target causal paths.
[0073] Step 407: Is the intervention finished?
[0074] Specifically, after determining the target causal path, determine whether the intervention is completed, that is, determine whether the incremental analysis intervention variable is still obtained. If the incremental analysis intervention variable is still obtained, then continue to return to execute step 404. If it is determined that the incremental analysis intervention variable is no longer obtained, then continue to execute step 408.
[0075] Step 408: Joint distribution update.
[0076] Specifically, each probability distribution corresponding to each target causal path is jointly updated to determine the joint distribution result of causal reasoning.
[0077] The advantage of this setting is that for the operation of incremental analysis intervention variables, the corresponding paths affected by the incremental analysis intervention variables are analyzed. Since the paths are independent of each other, we only need to focus on the dependencies between the variables in the affected paths. Through data dependency analysis, the affected variables are separated and the new subsequent probability distribution is updated. Through the above operations, the part that needs to be updated can be controlled to a minimum, so as to achieve the purpose of calculating only the affected path part and further updating the causal reasoning joint distribution result of the entire post-intervention distribution, thereby reducing the amount of reasoning calculation.
[0078] The present invention provides a causal reasoning method based on incremental analysis, which constructs a control causal graph; wherein the control causal graph contains all probability distributions corresponding to all causal variables and all candidate causal paths determined by all causal variables; obtains incremental analysis intervention variables; wherein the incremental analysis intervention variables refer to intervention variables that affect the candidate causal paths; determines all target causal paths according to the control causal graph and the incremental analysis intervention variables; wherein all target causal paths are paths affected by the incremental analysis intervention variables in all candidate causal paths in the control causal graph; updates each probability distribution corresponding to each target causal path, and determines the joint distribution result of causal reasoning. The technical solution of the present invention is used to solve the defect that the reasoning efficiency of the Bayesian causal reasoning method is restricted as the scale of variables increases in the prior art, and determines the affected target causal paths based on the candidate causal paths and the incremental analysis intervention variables, and only updates the probability distribution corresponding to the affected target causal paths, thereby improving the reasoning efficiency.
[0079] The causal reasoning device based on incremental analysis provided by the present invention is described below. The causal reasoning device based on incremental analysis described below and the causal reasoning method based on incremental analysis described above can be referenced to each other.
[0080] Figure 5 is a schematic diagram of the structure of the causal reasoning device based on incremental analysis provided by the present invention, referring to Figure 5 As shown, the causal reasoning device 500 based on incremental analysis includes: a construction module 501, a variable acquisition module 502, a path determination module 503 and a result determination module 504.
[0081] The construction module 501 is used to construct a control causal graph; wherein the control causal graph includes all probability distributions corresponding to all causal variables and all causal paths determined by all causal variables.
[0082] The variable acquisition module 502 is used to acquire incremental analysis intervention variables; wherein the incremental analysis intervention variables refer to intervention variables that affect the candidate causal path.
[0083] The path determination module 503 is used to determine all target causal paths according to the control causal graph and the incremental analysis intervention variables; wherein all target causal paths are the paths affected by the incremental analysis intervention variables among all candidate causal paths in the control causal graph.
[0084] The result determination module 504 is used to update each probability distribution corresponding to each target causal path respectively to determine the causal reasoning joint distribution result.
[0085] In an exemplary embodiment, the construction module 501 is specifically used to: obtain a causal graph; wherein the causal graph includes all causal variables and the dependencies between all causal variables; and construct a control causal graph according to the causal variables and dependencies in the causal graph.
[0086] In an exemplary embodiment, construction module 501 constructs a control causal graph based on causal variables and dependency relationships in the causal graph, and is specifically used to: determine an initial causal path based on the causal variables in the causal graph; determine a candidate causal path based on the dependency relationships between the initial causal path and all causal variables; obtain all initial probability distributions corresponding to all causal variables; and construct a control causal graph based on the candidate causal paths and all initial probability distributions.
[0087] In an exemplary embodiment, the path determination module 503 is specifically used to: determine whether the intervention variable branch corresponding to the intervention variable in the incremental analysis intervention variable satisfies the causal variable branch corresponding to the causal variable in all candidate causal paths in the control causal graph; when there is an intervention variable branch that does not satisfy the causal variable branch in all candidate causal paths, determine the candidate causal path corresponding to the causal variable branch that does not satisfy the target causal path; determine all target causal paths based on the candidate causal paths corresponding to the causal variable branches that do not satisfy the causal variable branches of all intervention variable branches.
[0088] In an exemplary embodiment, the path determination module 503 is further used to: when there is an intervention variable branch that satisfies the causal variable branch in all candidate causal paths, determine the candidate causal path corresponding to the causal variable branch that satisfies the causal variable branch as the remaining causal path; wherein the remaining causal path is the path among all candidate causal paths except all target causal paths.
[0089] In an example embodiment, the result determination module 504 is specifically used to: obtain the joint probability distribution corresponding to the remaining causal paths; wherein the joint probability distribution corresponding to the remaining causal paths remains unchanged; respectively update the probability distribution corresponding to each target causal path to determine the target probability distribution corresponding to each target causal path; and determine the causal reasoning joint distribution result based on the sum of all target probability distributions and the joint probability distribution.
[0090] The device of this embodiment can be used to execute the method of any embodiment in the embodiment of the causal reasoning method based on incremental analysis. Its specific implementation process and technical effects are similar to those in the embodiment of the causal reasoning method based on incremental analysis. For details, please refer to the detailed introduction in the embodiment of the causal reasoning method based on incremental analysis, which will not be repeated here.
[0091] Figure 6 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the causal reasoning method based on incremental analysis, which includes: constructing a control causal graph; wherein the control causal graph contains all probability distributions corresponding to all causal variables and all candidate causal paths determined by all causal variables; obtaining incremental analysis intervention variables; wherein the incremental analysis intervention variables refer to intervention variables that affect the candidate causal paths; determining all target causal paths according to the control causal graph and the incremental analysis intervention variables; wherein all target causal paths are paths in all candidate causal paths in the control causal graph that are affected by the incremental analysis intervention variables; respectively updating each probability distribution corresponding to each target causal path to determine the joint distribution result of causal reasoning.
[0092] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0093] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the causal reasoning method based on incremental analysis provided by the above methods, and the method includes: constructing a control causal graph; wherein the control causal graph contains all probability distributions corresponding to all causal variables and all candidate causal paths determined by all causal variables; obtaining incremental analysis intervention variables; wherein the incremental analysis intervention variables refer to intervention variables that affect the candidate causal paths; determining all target causal paths based on the control causal graph and the incremental analysis intervention variables; wherein all target causal paths are paths among all candidate causal paths in the control causal graph that are affected by the incremental analysis intervention variables; and updating each probability distribution corresponding to each target causal path respectively to determine the joint distribution result of causal reasoning.
[0094] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the causal reasoning method based on incremental analysis provided by the above-mentioned methods, the method comprising: constructing a control causal graph; wherein the control causal graph contains all probability distributions corresponding to all causal variables and all candidate causal paths determined by all causal variables; obtaining incremental analysis intervention variables; wherein the incremental analysis intervention variables refer to intervention variables that affect the candidate causal paths; determining all target causal paths based on the control causal graph and the incremental analysis intervention variables; wherein all target causal paths are paths among all candidate causal paths in the control causal graph that are affected by the incremental analysis intervention variables; and updating each probability distribution corresponding to each target causal path respectively to determine the joint distribution result of causal reasoning.
[0095] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0096] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A causal reasoning method based on incremental analysis, characterized in that: include: Constructing a control causal graph; wherein the control causal graph includes all probability distributions corresponding to all causal variables and all candidate causal paths determined by all the causal variables; Obtaining an incremental analysis intervention variable; wherein the incremental analysis intervention variable refers to an intervention variable that affects the candidate causal path; Determine all target causal paths according to the control causal graph and the incremental analysis intervention variables; wherein all the target causal paths are paths among all the candidate causal paths in the control causal graph that are affected by the incremental analysis intervention variables; The probability distributions corresponding to the target causal paths are updated respectively to determine the joint distribution result of causal reasoning.
2. The causal reasoning method based on incremental analysis according to claim 1 is characterized in that: The construction of the control cause-effect diagram comprises: Obtaining a causal graph; wherein the causal graph includes all the causal variables and the dependency relationships between all the causal variables; The control causal graph is constructed according to the causal variables and the dependency relationships in the causal graph.
3. The causal reasoning method based on incremental analysis according to claim 2 is characterized in that: The step of constructing the control causal graph according to the causal variables and the dependency relationships in the causal graph includes: determining an initial causal path according to the causal variables in the causal graph; Determine the candidate causal path according to the dependency relationship between the initial causal path and all the causal variables; Obtain all initial probability distributions corresponding to all the causal variables; The control causal graph is constructed according to the candidate causal paths and all the initial probability distributions.
4. The causal reasoning method based on incremental analysis according to claim 1 is characterized in that: Determining all target causal paths according to the control causal diagram and the incremental analysis intervention variables includes: Determine whether the intervention variable branch corresponding to the intervention variable in the incremental analysis intervention variable satisfies the causal variable branches corresponding to the causal variables in all the candidate causal paths in the control causal graph; In the case that there is an intervening variable branch that does not satisfy the causal variable branch in all the candidate causal paths, determining the candidate causal path corresponding to the branch that does not satisfy the causal variable branch as the target causal path; All the target causal paths are determined according to the candidate causal paths corresponding to the causal variable branches that all the intervention variable branches do not satisfy.
5. The causal reasoning method based on incremental analysis according to claim 4 is characterized in that: Also includes: In the case that there is an intervening variable branch that satisfies the causal variable branch in all the candidate causal paths, the candidate causal path corresponding to the causal variable branch is determined as the remaining causal path; wherein the remaining causal path is the path among all the candidate causal paths except all the target causal paths.
6. The causal reasoning method based on incremental analysis according to claim 5 is characterized in that: The updating of the probability distributions corresponding to the target causal paths respectively to determine the joint distribution result of causal reasoning includes: Obtaining the joint probability distribution corresponding to the remaining causal paths; wherein the joint probability distribution corresponding to the remaining causal paths remains unchanged; Probability updating is performed on each of the probability distributions corresponding to each of the target causal paths to determine the target probability distribution corresponding to each of the target causal paths; The causal inference joint distribution result is determined according to the sum of all the target probability distributions and the joint probability distribution.
7. A causal reasoning device based on incremental analysis, characterized in that: include: A construction module, used to construct a control causal graph; wherein the control causal graph includes all probability distributions corresponding to all causal variables and all causal paths determined by all the causal variables; A variable acquisition module, used to acquire an incremental analysis intervention variable; wherein the incremental analysis intervention variable refers to an intervention variable that affects the candidate causal path; A path determination module, used to determine all target causal paths according to the control causal graph and the incremental analysis intervention variables; wherein all the target causal paths are paths affected by the incremental analysis intervention variables among all the candidate causal paths in the control causal graph; The result determination module is used to update the probability distributions corresponding to the target causal paths respectively to determine the joint distribution results of causal reasoning.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the causal reasoning method based on incremental analysis as described in any one of claims 1 to 6 is implemented.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the causal reasoning method based on incremental analysis as described in any one of claims 1 to 6 is implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the causal reasoning method based on incremental analysis as described in any one of claims 1 to 6 is implemented.