30 results about "Topological order" patented technology

The invention discloses a vehicle-mounted controller firmware upgrading method, which relates to the technical field of intelligent networked automobiles, and comprises the following steps of: accurately describing a dependency relationship between versions by introducing a directed acyclic graph, providing a clear topological sorting basis for an upgrading process, and improving the upgrading efficiency. The system function conflict and disorder caused by disordered upgrading sequence are fundamentally avoided; a distributed consensus protocol is utilized to ensure that each controller achieves state coordination before upgrading, so that the overall robustness of a multi-controller cluster is effectively improved when the multi-controller cluster faces network delay or individual node abnormity, and upgrading failure or version bifurcation caused by the fact that part of nodes are not ready is avoided; through the combination of the group-level binding token and the atomic switching mechanism, the controller group with a strong function dependency relationship can be ensured to be effective and updated as a unified transaction unit, and the safety and reliability of cross-domain function linkage are remarkably enhanced.

A system and method include learning a topological order of a plurality of variables in a directed acyclic graph based on real data, computing parameter estimate values corresponding to the real data, computing error values based on the real data, the topological order, and the parameter estimate values, generating simulated data from the parameter estimate values and the error values, such that simulated variables in the simulated data preserve a causal relationship between variables in the real data, and the simulated variables in the simulated data preserve a correlation relationship between the variables in the real data, and reorganizing and outputting the simulated data based on the topological order.

The invention discloses an assessment method and device for a large model generated text, a readable storage medium and a program product, and relates to the technical field of artificial intelligence, and the assessment method for the large model generated text comprises the steps: obtaining a user-defined assessment template, and analyzing a dependency relationship and an execution priority of each preset assessment dimension in the user-defined assessment template, obtaining a directed acyclic graph used for representing the topological sequence of each preset evaluation dimension; according to a topological sequence of each preset evaluation dimension in the directed acyclic graph, evaluating the to-be-evaluated large model generation text on each preset evaluation dimension to obtain a score corresponding to each preset evaluation dimension; and determining a text evaluation result by counting the score corresponding to each preset evaluation dimension. According to the technical scheme, multi-dimensional evaluation can be carried out on the large model generation text, the multi-aspect capability of the model can be measured more comprehensively and accurately, and the scene adaptation capability of the text evaluation scheme is improved.

A system and method include learning a topological order of a DAG by setting an initial index value of a first index, setting an initial score value of a score, setting an initial order list, computing an initial SSCP matrix, sweeping the initial SSCP matrix based on the first index, incrementing the first index to obtain an updated index value of the first index, determining an index value of a second index, computing an updated SSCP matrix, computing an updated score value as a sum of an initial score value and a value identified from the updated SSCP matrix, and computing an updated order list from an initial order list based on the updated index value of the first index and the index value of the second index.

The invention discloses a code abstract generation method and system based on hierarchical context awareness, and belongs to the field of natural language processing and software engineering. The method comprises the following steps: generating hierarchical annotations for a code warehouse based on a function call graph, firstly extracting a call relationship in the graph and eliminating a ring structure, and then classifying leaf nodes and generating annotations; a topological sequence from bottom to top is adopted in the core step, the importance score of a child node is calculated for each father node, child node annotations, class function descriptions and code bodies of the child nodes are fused, and a large language model is input to generate annotations till the whole graph is covered. The method is suitable for an automatic document generation scene of a complex software project, and through a technical path combining function call graph analysis, child node importance measurement and class function description, the accuracy and semantic consistency of code annotation are improved; the method can be widely applied to the fields of software maintenance, code understanding, intelligent development tools, automatic document generation of enterprise-level code warehouses and the like.

The application provides a causal discovery method and device fusing field knowledge of a large language model and a flow model, and relates to the technical fields of artificial intelligence and data mining. The method comprises the following steps: acquiring observation data and variable semantic information of a system to be analyzed; determining a causal topological order between variables by using a large language model; constructing a mask affine autoregressive flow model based on the causal topological order, and training the model by using the observation data; calculating a Jacobian matrix of the flow model with respect to input data, constructing a statistic based on the Jacobian matrix, and performing hypothesis testing; pruning according to a test result to obtain a candidate causal graph; inputting the candidate causal graph into the large language model, and performing secondary verification and pruning on candidate connections by using the variable semantic information to obtain a final sparse causal graph. The application introduces prior knowledge by using semantic information of the large language model, and fits data by using a more flexible function form of the flow model, thereby significantly improving the accuracy and robustness of causal discovery.

The invention discloses a calculation optimization method and device, a storage medium and a computer program product, and belongs to the technical field of computers. The method comprises the following steps: determining a topological sequence of a computational graph; determining a target operator node in the plurality of operator nodes and a target input node corresponding to the target operator node based on a calculation relationship represented by the topological sequence and the calculation graph, the target operator node being an operator node used for in-situ calculation in the calculation graph, the memory used by the output data of the target operator node is consistent with that used by the input data of the target input node; and executing a calculation task corresponding to the calculation graph based on the target operator node and the target input node. That is, according to the scheme, the nodes capable of performing in-situ calculation in the calculation graph can be automatically identified by the computer equipment. Compared with a method of manually assigning in-situ calculation, the scheme has the advantages that errors are not prone to occurring, efficiency is higher, most of even all nodes capable of conducting in-situ calculation can be found, and therefore the overall calculation performance can be greatly improved.

The application provides a causal discovery method and device fusing field knowledge of a large language model and a flow model, and relates to the technical fields of artificial intelligence and data mining. The method comprises the following steps: acquiring observation data and variable semantic information of a system to be analyzed; determining a causal topological order between variables by using a large language model; constructing a mask affine autoregressive flow model based on the causal topological order, and training the model by using the observation data; calculating a Jacobian matrix of the flow model with respect to input data, constructing a statistic based on the Jacobian matrix, and performing hypothesis testing; pruning according to a test result to obtain a candidate causal graph; inputting the candidate causal graph into the large language model, and performing secondary verification and pruning on candidate connections by using the variable semantic information to obtain a final sparse causal graph. The application introduces prior knowledge by using semantic information of the large language model, and fits data by using a more flexible function form of the flow model, thereby significantly improving the accuracy and robustness of causal discovery.

A system and method include generating a topological order of a DAG by creating residual series vectors, calculating normality statistic and MSE values for the residual series vectors, comparing the normality statistic values with a critical value, for each normality statistic value that is less than or equal to the critical value, adding a variable index to a temporary order list and the MSE value to an MSE list, counting a number of elements in the temporary order list, if the number of elements in the temporary order list is zero, updating an order list based on the normality statistic values or if the number of elements in the temporary order list is not zero, updating the order list based on at least one of the temporary order list or the MSE list, and outputting the order list as the topological order of the DAG.

The invention belongs to the technical field of artificial intelligence and natural language processing, and particularly relates to a large model multi-hop tool calling training data construction method based on subgraph backtracking, which comprises the following steps: acquiring a large model callable tool to obtain a tool name, a tool description and input and output parameter information; constructing a tool graph which takes tools as nodes and dependencies between input and output of the tools as directed edges; sampling directed acyclic tool sub-graphs with different tool numbers from the constructed tool graph; traversing all dependent edges in the tool sub-graph according to a sub-graph topological sequence, and generating sub-task data corresponding to the dependency relationship by utilizing a first large language model based on tool information connected with each edge; synthesizing all the generated sub-tasks into a complete natural language query by using a second large language model; and verifying the synthesized data and the corresponding tool execution track to generate high-quality multi-hop tool learning data.

The application discloses a kind of vehicle-mounted controller firmware upgrade method, it is related to intelligent network connection automobile technical field, including, the application is accurately described the dependence between versions by introducing directed acyclic graph, provides clear topological order basis for upgrade process, fundamentally eliminates the system function conflict and disorder caused by upgrade sequence confusion;Utilize distributed consensus protocol to ensure that each controller reaches state coordination before upgrading, effectively improves the overall robustness of multi-controller cluster when facing network delay or individual node exception, avoids upgrade failure or version bifurcation caused by part of node not ready;Through group-level binding token and atomic switching mechanism, it is guaranteed that the controller group with strong dependence relationship can be updated as a unified transaction unit, significantly enhancing the security and reliability of cross-domain function linkage.

Non-Abelian anyons are created, braided, and fused using the physical qubits of a QCCD-based quantum processor. A controller controls operation of a confinement apparatus to cause a plurality of physical qubits to be confined by the confinement apparatus. At least some of the plurality of physical qubits are logically organized onto a lattice and have been prepared to provide a non-Abelian topological order ground state. The lattice is formed of a plurality of sublattices. The controller causes performance of an anyon creation gate to cause creation of a first pair of non-Abelian anyons on a sublattice of the plurality of sublattices; causes a path traversal gate sequence to be performed to cause at least a first anyon of the first pair of anyons to traverse a braiding path to form a closed loop on the sublattice; and determines a fusion channel formed by fusing the first pair of anyons.

This application provides an automated code repository generation method, electronic device, and program product, relating to the field of software design and development. The method includes: acquiring functional specification information; generating corresponding functional subtrees based on the functional specification information; clustering nodes in the functional subtrees into multiple functional modules based on the cohesion principle, and generating a modular functional diagram; mapping the functional modules in the modular functional diagram to file structures, generating an enhanced file structure diagram; modeling data flows between functional modules and file sequence dependencies within functional modules based on the enhanced file structure diagram to generate a repository planning diagram; and generating code files using a test-driven development process based on the topological order determined by execution dependencies to form a code repository. This method enhances consistency assurance from high-level functional planning to low-level code implementation, improving the modular quality and maintainability of the code repository generation process.

This invention discloses a fund clearing method based on dynamically constructed directed acyclic graphs (DAGs). The method includes: receiving a transaction request containing transaction context information; selecting a clearing strategy based on this information and instantiating a clearing strategy DAG in memory; traversing and executing the rule templates corresponding to each node according to the topological order of the DAG; collecting atomic accounting instructions generated by all nodes during the traversal; and submitting the instructions to the core accounting system for execution after the traversal is complete. This invention, by deconstructing the clearing logic into a dynamically constructed DAG, achieves separation of business logic and execution engine, improving business flexibility and system performance. It can be used to handle complex transaction scenarios with high concurrency and changing rules, while simplifying auditing and reducing fund processing risks.

The application discloses a campus safety risk identification method and system based on multi-algorithm nested cooperation and a medium, relates to the technical field of campus risk identification, and constructs a dynamic rule library, matches all dynamic rule graphs corresponding to initial event signals from the dynamic rule library, and creates dynamic rule graph instances; each node is activated according to a topological order and an activation condition, the comprehensive confidence and the quasi-truth degree of a potential risk hypothesis are calculated at a convergence node based on DS evidence theory to perform risk early warning; multi-dimensional information is deeply associated and comprehensively judged, and the problems of information silos and low-level misjudgment are solved; a school calendar rhythm is converted into a time weight factor and embedded into a DS evidence fusion process, time background can be perceived, the contribution degree of different types of evidence and the sensitivity of risk decision-making can be automatically adjusted, false positives caused by neglecting context are reduced, and early and accurate early warning can be ensured in a high-risk period.

The invention discloses an indirect prompt injection defense method and system for a large language model agent, and belongs to the field of artificial intelligence safety. Analyzing a user task instruction and predefined tool description based on a large language model agent, and generating a directed acyclic graph of a tool dependency relationship; tool calling of each node is executed according to a topological sequence, and a parameter reasoning result generated by node response is written into a context; if the information needs to be supplemented, allowing the agent to safely expand the node; when a potential injection instruction is detected, external tools related to semantics are guided to be called and added into a pseudo tool set, and finally the set is processed through a pseudo tool calling mechanism to achieve active defense. According to the method, the task execution process is explicitly modeled into the directed acyclic graph representing the tool dependency relationship, the tool calling path is predefined in the task planning stage, the behavior of the agent is constrained in the task execution process, and the illegal tool calling behavior triggered by the injection instruction is inhibited from the source.

Non-Abelian anyons are created, braided, and fused using the physical qubits of a QCCD-based quantum processor. A controller controls operation of a confinement apparatus to cause a plurality of physical qubits to be confined by the confinement apparatus. At least some of the plurality of physical qubits are logically organized onto a lattice and have been prepared to provide a non-Abelian topological order ground state. The lattice is formed of a plurality of sublattices. The controller causes performance of an anyon creation gate to cause creation of a first pair of non-Abelian anyons on a sublattice of the plurality of sublattices; causes a path traversal gate sequence to be performed to cause at least a first anyon of the first pair of anyons to traverse a braiding path to form a closed loop on the sublattice; and determines a fusion channel formed by fusing the first pair of anyons.

Non-Abelian anyons are created, braided, and fused using the physical qubits of a QCCD-based quantum processor. A controller controls operation of a confinement apparatus to cause a plurality of physical qubits to be confined by the confinement apparatus. At least some of the plurality of physical qubits are logically organized onto a lattice and have been prepared to provide a non-Abelian topological order ground state. The lattice is formed of a plurality of sublattices. The controller causes performance of an anyon creation gate to cause creation of a first pair of non-Abelian anyons on a sublattice of the plurality of sublattices; causes a path traversal gate sequence to be performed to cause at least a first anyon of the first pair of anyons to traverse a braiding path to form a closed loop on the sublattice; and determines a fusion channel formed by fusing the first pair of anyons.

The invention belongs to the technical field of image recognition, and particularly relates to an intelligent damage detection and evaluation method and system for an accident vehicle, and the method comprises the steps: constructing a normalized distortion energy diagram of a local damage image of the accident vehicle, and initializing a particle swarm to form an annular topological structure; performing iterative search by adopting a particle swarm optimization algorithm, in each iteration, calculating an inertia weight according to the distortion energy value at the current position of the particle, and updating the velocity vector and the position of the particle according to the inertia weight and the distortion energy value at the position of the particle in the topological neighbor set of the particle, and connecting the particles according to a topological sequence when iteration stops to obtain a damage closed polygon contour, further calculating a physical area and a depth index of a damage region, and generating a damage assessment report. According to the invention, automatic repair and evaluation of the damaged edge of the accident vehicle are realized, subjective errors of manual loss assessment are eliminated, and objectivity and accuracy of loss assessment and claim settlement of the vehicle insurance are improved.

The invention discloses an anti-side channel attack circuit compiling method based on unique random protection, and belongs to the field of cryptographic engineering and hardware security. In order to solve the problem of contradiction between low random consumption and multiple rounds of safety certification of the existing mask scheme, a robust detection safety model and a unique random protection criterion are constructed; initial masking is carried out on circuit main input, linear and nonlinear gates are converted according to a topological sequence, protection collision detection and correction, module output re-masking and random number recovery and safe reutilization are executed, and an overall safe combination structure is constructed. The method can systematically generate a mask circuit which has provable first-order security and can resist glitch attacks under the condition of only depending on the initial random entropy.

The invention relates to the field of air pollution intelligent analysis and multi-agent collaboration, and provides an air pollution sub-agent automatic arrangement method and system based on business semantic ontology. The method comprises the following steps: encoding an input natural language question request, decoding to obtain a service intention under the constraint of a six-domain ontology, and calculating a tetrad confidence coefficient; if the confidence coefficient of the tetrad is greater than a set threshold value, performing space-time intersection filtering and root node matching based on a service intention to obtain an empty node template; instantiating the business intention into a plurality of sub-agent nodes by taking the empty node template as a root, constructing a directed acyclic graph, and performing space-time intersection pruning to obtain an instantiated topological structure; injecting real-time data into the instantiated topological structure, and writing user preferences to obtain an instantiated directed acyclic graph; and on the basis of the instantiated directed acyclic graph, asynchronously scheduling each sub-agent according to a topological sequence through a Redis Stream message bus.

The invention relates to a neural network accelerator, in particular to a memory allocation method for a multi-level cache neural network accelerator, which comprises the following steps of: reordering neural network calculation graphs according to a topological sequence according to a calculation dependency relationship among operators; according to a topological sequence, respectively collecting the input tensor of each operator and the cache size required by the output tensor in the L2buffer and the L3buffer; according to the topological sequence and the calculation characteristics of each operator, the multiplexing condition of the input tensor in the execution process of each operator is collected; collecting condition information of each operator according to the topology sequence and the characteristics of each operator; calculating the life cycles of all operator tensors according to the topological sequence of the operators, and calculating the use intensity of all operator tensors; sorting all operator tensors based on the use intensity, and starting to distribute from the operator tensor with the maximum use intensity; the method can effectively overcome the defect that the inference efficiency of the neural network is low due to unreasonable memory allocation and tensor arrangement.

The application provides a kind of fuzzy testing method based on symbolic execution, and the fuzzy testing tool first starts symbolic execution and analysis program;The symbolic execution and analysis program is constructed on binary level;Symbolic execution and analysis program call stain analysis program to carry out binary symbol-based stain analysis after receiving request;Stain analysis program generates the record file of contaminated instruction after completing stain analysis, and stores record file in the symbolic execution and analysis program under the output directory of fuzzy testing tool;Symbolic execution and analysis program analyzes record file to obtain field and data block information by heuristic algorithm to identify record file structure, and constructs operand dependency graph;Symbolic execution and analysis program traverses operand dependency graph according to inverse topological order, and responds to fuzzy testing tool after completing hybrid symbolic execution by establishing symbolic expression for each binary instruction.The application reduces the loss of information in the process of stain analysis and symbolic execution, has good efficiency and high accuracy.

A method for detecting topological order in a quantum state of a quantum computer (qc) comprising a plurality of qubits (qb) is specified herein, the method comprising the steps of: - performing at least a partial quantum error correction on the quantum state using a global decoder, wherein a syndrome (syn) configuration of the quantum state is determined by measuring local stabilizer operators, - measuring a 1-form symmetry operator (op) on the at least partially quantum error corrected quantum state, and - detecting topological order in the quantum state by measuring at least one open operator (oo) associated with the 1-form symmetry operator (op) on the at least partially quantum error corrected quantum state. Further, a quantum computer is specified herein.

The invention discloses an SIMD architecture-oriented neural network processor intra-core scheduling method and system, and belongs to the technical field of processor scheduling. Through a three-level screening mechanism of the node type priority, the time window priority and the memory influence degree priority, in combination with dependency constraint processing and L0 cache mutual exclusion constraint management, a scheduling sequence which meets a topological sequence and has the minimum peak cache residence amount is generated, idle computing power in a chip is effectively utilized, and the resource utilization rate is improved. On the basis, cache allocation and swap-in and swap-out operation are optimized based on a cost-aware strategy, an optimized scheduling sequence containing swap-in and swap-out operation nodes is generated, swap-in and swap-out memory minimization is achieved, and the scheduling efficiency and the resource utilization rate are further improved. In addition, the method further optimizes the swap-in and swap-out position, and further reduces the total operation time of the operator on the premise of ensuring that the data carrying amount except the total amount is not obviously increased.

The invention provides a feature analysis method and device, electronic equipment and a storage medium, and belongs to the technical field of artificial intelligence, and the method comprises the steps: obtaining meta-information of each to-be-analyzed feature; the meta-information comprises a feature name, an analyzer and a dependency parameter; combining all the same meta-information of the resolver; establishing a feature dependence tree containing each piece of meta-information; the feature name of the father node in the feature dependency tree corresponds to the dependency parameter of the child node; and based on the topological sequence of the feature dependency tree, calling the parser of each node to obtain the feature corresponding to each node. According to the method, the information of the same feature elements of the resolvers is combined, the more simplified feature dependence tree is established, the resolvers of the nodes are called based on the topological sequence of the feature dependence tree, the total calling frequency of the resolvers is reduced, and therefore time consumed in the feature obtaining stage can be reduced.

In one aspect, a method for determining personalized search results with respect to mandatory constraints and qualitative preferences on search criteria representable via a DAG includes collecting input from users or software systems; transforming the inputs into the mandatory constraints and preferences on search criteria, wherein the search criteria is partially ordered and representable through the DAG; determining weights for the search criteria using a function, and associating the weights with the nodes of the DAG to determine a partial ordering of such weights equivalent to an order of the DAG nodes, where any node at level “k” has a weight “w” such that sum of all or any finite subset of the weights of the nodes at levels below “k” is always less than “w”, the level being defined by a topological order of the DAG; and determining search results that satisfy the mandatory constraints on the search criteria.

The application provides a kind of based on heterogeneous collaborative data node quantification calculation method and system, belong to computer data processing and collaborative work analysis technical field, in view of the problems that heterogeneous collaborative data is difficult to be unified modeling in prior art, work achievement correlation is difficult to be stably identified, direct contribution and indirect contribution are difficult to be unified quantification, the original event data of at least two heterogeneous collaborative platforms is acquired, work achievement node is generated after preprocessing, candidate associated edge is generated according to preset correlation rule, and directed acyclic contribution correlation graph is constructed after graph constraint processing;Identify terminal achievement node and distribute initial node quantification value, perform reverse propagation calculation along the inverse topological order of contribution correlation graph, obtain node quantification value and subject aggregation quantification value.The application can improve the fusion capability of heterogeneous collaborative data, the stability of correlation identification and the traceability of quantification result.

The invention provides a data query method, electronic equipment and a storage medium, and relates to the field of data query. Constructing a directed acyclic graph according to data query information input by a user; nodes of the directed acyclic graph correspond to tool execution units. According to the topological sequence of the directed acyclic graph, tool execution units corresponding to all the nodes are triggered in sequence to execute corresponding operations till a database query instruction is obtained. The input parameters are verified before the tool execution unit executes the corresponding operation, so that the target input parameters are obtained according to the verification result, and the reliability of the data query instruction is improved. At least one of the tool identifier of the tool called by executing the corresponding operation, the target input parameter and the execution result is returned to the client, so that a user can conveniently judge the logic rationality of generating the database query instruction, the necessity of tool calling and the accuracy of each execution result, the process of generating the database query instruction is transparent, and the user experience is improved. The credibility of the user on the target data is improved, and the anomaly analysis difficulty is reduced.

The application provides a method and system for bidirectional data type derivation in graphical programming, and relates to the technical field of graphical programming, and comprises the following steps: S1, parsing graphical programming data into a graph structure and a tree structure; S2, performing data type derivation from outside to inside: based on the tree structure, traversing each graph structure in a depth-first order, for a currently traversed graph structure, traversing internal nodes in a topological order of a directed acyclic graph, and according to a data flow direction, deriving a data type of a downstream node according to a data type of an upstream node; S3, performing data type derivation from inside to outside: based on the tree structure, traversing each graph structure in a postorder, for a currently traversed graph structure, determining a data type of an output boundary according to a data type of an internal node, and reversely propagating the data type to a corresponding node in a parent graph; and S4, automatically generating type conversion code according to a predefined data type processing rule.