77results about How to "Improve reasoning efficiency" patented technology

The invention discloses an intelligent fault diagnosis system for an ICNI system, which can improve the maintenance efficiency, carry out intelligent and automatic diagnosis and is applicable to the ICNI system. According to the technical scheme of the invention, a knowledge base and a management module thereof carry out standardization research and mathematical modeling based on a fault tree, an SQL Server database software framework is adopted, a relational database is used for building a logic relation among a fault phenomenon, a fault mode, a detection method, a historical case and a fault tree internal event to form the knowledge base; and a diagnosis information acquisition module interacts with an automatic testing system via Ethernet to acquire diagnosis data from the ICNI system and a testing instrument, a reasoning machine module adopts CBR and RBR hybrid diagnostic reasoning, after comprehensive judgment is carried out on the fault phenomenon inputted by the user, the field knowledge stored by the knowledge base and the diagnosis data from the automatic testing system, a reasoning method is automatically selected to carry out reasoning diagnosis on the fault, a reasoning process and a reasoning result are outputted to an explanation machine module, and a diagnosis report is generated.

An efficient reasoning system and method using a view in a DBMS-based RDF triple store are provided. The DBMS-based reasoning system includes a triple input unit for receiving a Resource Description Framework (RDF) triple. A triple examination unit examines whether the received triple conforms to RDFS subsumption relation entailment rules or Web Ontology Language (OWL) inverse relation rules. A view creation unit creates a table view when the received triple conforms to the RDFS subsumption relation entailment rules or the OWL inverse relation rules as a result of the examination. A triple storage unit stores the received triple. The DBMS-based triple store can efficiently perform reasoning based on rule rdfs7 or rdfs9, which is included in the RDFS subsumption relation entailment rules, and the OWL inverse relation rules.

The invention relates to a self-learning mechanism-base fast matching fuzzy reasoning method. The method includes the following steps that: a Gaussian membership degree function method is adopted to construct parameter fuzzification information; a fuzzy rule base is established; external parameters are fuzzificated, so that a fact item can be obtained; the fact item is matched with rules in the fuzzy rule base by adopting a rete algorithm, so that a fuzzy reasoning result can be obtained; the fuzzy reasoning result is subjected to defuzzification, so that a final reasoning result can be obtained; and a sample set is constructed according to the final reasoning result and an actual feedback result, and rule strength self-learning correction is carried out based on the sample set. According to the self-learning mechanism-base fast matching fuzzy reasoning method of the invention, the rete algorithm is adopted, so that the efficiency of fuzzy reasoning can be improved, and the fuzzy reasoning method can be applied to the engineering field with high real-time requirements.

The invention provides a forward reasoning method and device for neural network, equipment and a storage medium The method comprises the following steps of: dividing the target neural network into a plurality of sub-networks, wherein any one sub-network comprises at least one hidden layer of the target neural network, creating inference instances and inference engines corresponding to the plurality of sub-networks respectively on hardware equipment of the inference platform, and performing forward inference on the target neural network based on the inference instances and inference engines corresponding to the plurality of sub-networks respectively. One inference engine is only responsible for a part of hidden layers of the neural network, and multiple data inputs can be executed in different inference engines in parallel at the same time, so that the forward inference method provided by the invention has relatively high inference efficiency and data throughput, and hardware resourcesof an inference platform are fully utilized.

The invention relates to a meeting notice system and a method based on context service. The system is provided with five components: a client side, a service bus, a BPEL execution engine, a context service module and a communication service module. The meeting notice method comprises the following steps: after inputting a meeting notice request from the client side by a user, receiving the requestby the service bus, transmitting the request to the BPEL execution engine, and starting a meeting notice process module by the BPEL execution engine; in allusion to every conventioneer, respectivelyexecuting meeting notice service by the meeting notice process module: firstly, calling the context service module to obtain how to notify the current conventioneers and contact ways thereof and notice contents, then calling a corresponding service unit in the communication service module by the service bus in order to send a meeting notice to the conventioneer, and returning a notice result to the meeting notice process module; judging whether to notify the next conventioneer or not by the module, and returning a final result to the client side after notifying all conventioneers. The invention uses a plurality of communication means to automatically complete the meeting notice service through combining a semanteme reasoning technique.

The invention discloses a breakdown diagnosis method for semiconductor devices. The method includes the following steps: firstly establishing an expert system which includes a fuzzy rule database and an inference engine; further, conducting fuzzy processing on a semiconductor device real-time monitoring parameter to generate a fuzzy fact; the inference engine, on the basis of the fuzzy fact, performing fuzzy processing by interacting with the fuzzy rule database to generate a breakdown diagnosis result; and on the basis of actual application conditions of the breakdown diagnosis result, carrying out rule intensity self-learning correction targeted at the fuzzy rule database. According to the invention, the method performs fuzzy inference and further obtains the breakdown diagnosis result based on the rete algorithm, and carries out rule intensity self-learning correction on the fuzzy rule database in combination with the actual application of the breakdown diagnosis result, so that the method increases the precision and automation of the breakdown diagnosis for semiconductor devices and ensures efficient and stable operation of the semiconductor technology.

The invention relates to a method for constructing a highly shared Rete network, which belongs to the fields of artificial intelligence and expert systems. The invention mainly studies the rule reasoning technology in the expert system. On the basis of the original Rete reasoning technology, a high sharing Rete network construction algorithm based on the node sharing degree and mode sharing degree model is proposed. The invention improves the node sharing performance of the Rete network, reduces redundant nodes, optimizes the structure of the Rete network nodes, and can significantly improve the efficiency of rule reasoning.

The invention discloses a task guidance type smart agricultural planting expert system. A knowledge obtaining module builds a knowledge base by obtaining agricultural planting knowledge, a knowledge base management module sets confidence for each rule type knowledge, and the inference accuracy of uncertain events is improved; The reasoning task is generated; An improved Rte network mode matching algorithm is adopted; knowledge stored in the knowledge base is inquired and a rule required by the reasoning task is subjected to matching and activating operation in the knowledge base and variablesin a variable library and the external user data are combined, reasoning of the reasoning task is completed according to the confidence coefficient of the activated rule, and finally a structured production task is generated for guiding a user to carry out agricultural production activities to realize scientific planting. By adopting the embodiment of the invention, the method can be extended to model establishment and reasoning of other crops, and after a rule base and a knowledge base are continuously huge, the method has higher matching speed, reasoning efficiency and model generalization.

The invention discloses a method for constructing a cascading Bayesian network for solving a combinatorial explosion problem. The method comprises two core parts: A) a method for constructing the topological structure of a cascading Bayesian network by a system access, and B) an intermediate node probability parameter setting method of the cascading Bayesian network. Firstly, the construction of the topological structure of the Bayesian network is finished, wherein the topological structure of the Bayesian network is consistent with a fault cascading configuration in engineering practice, then, the setting of each node condition probability table of the Bayesian network is finished, and finally, the construction of the cascading Bayesian network is jointly finished. When the cascading Bayesian network finishes being constructed, any existing Bayesian network reasoning technology can be applied to carry out reasoning calculation on the cascading Bayesian network so as to obtain system reliability. While a reliability calculation result of the system can be guaranteed to be correct, the probability parameters in the network are effectively reduced, specifically, the number of the probability parameters on each access can be lowered to a linear order from an exponential order, calculation efficiency is improved, and a combinatorial explosion problem is solved.

The invention discloses an OWLHorst rule distributed type parallel reasoning algorithm in combination with a Spark platform. According to the characteristics of Spark RDD, the principle of a TREAT platform is combined, an alpha register Om_RDD or Pt_RDD corresponding to a mode triad is constructed for RDF ontology data and broadcast, and a rule marking model is constructed; a mode first component of each rule is connected, a corresponding connecting mode triad set Rulem_linkvar_RDD is generated, and therefore the matching speed in the reasoning process is increased. At the OWL Horst reasoning stage, an alpha stage in a TREAT algorithm is achieved in combination with MapReduce, distributed parallel reasoning of multiple rules is achieved, and then the reasoning result is subjected to de-weight processing; a large number of instance triads can be filtered through the alpha register and the rule marking model, output of key assignment pairs at a Map stage is reduced, and therefore invalid network transmission is reduced.

Embodiments of the invention disclose a context distributed reasoning method and device. The method comprises the following steps: obtaining mass of context data in various types from a plurality of computing nodes, and carrying out modeling processing on the obtained context data to obtain a reasoned context set, wherein the expression forms of various contexts in the reasoned context set are uniform and the mass indicates that the scale of the context data achieves a specific data amount; obtaining a reasoning rule which is used for carrying out distributed reasoning on the reasoned context set; analyzing the reasoning rule to generate a reasoning plan of the reasoning rule; carrying out distributed reasoning on the reasoned context set according to the reasoning plan and a pre-configured context reasoning algorithm for the reasoning rule so as to generate a reasoning result of the reasoning plan. According to the method and device disclosed in the invention, the context reasoning efficiency can be improved.

The invention relates to a network event correlation analysis and dynamic early-warning method. The method is characterized in that a network event is subjected to correlation analysis pretreatment to form an alarm event according to the requirement of the network fault management, a priority reasoning method is selected by a reasoning method selector, the alarm event is treated by the priority reasoning method and a lower priority reasoning method sequentially until the treatment is successful; if the two reasoning methods fail in treatment, dynamic early-warning is provided to network operation and maintenance staff. The method can select the appropriate priority reasoning method in a self-adaption manner, the reasoning efficiency is greatly improved, the case library and the rule base can be optimized constantly in a network event treatment process and the technical requirements for the network operation and maintenance staff are reduced.

The invention discloses an intelligent vehicle fault reasoning method based on a Bayesian network, and the method comprises the steps: building a fault tree model based on the Bayesian network, and obtaining the prior probability of a fault tree node and a detection method of an associated fault tree node; inputting the fault tree model, the prior probability and the detection method into an inference engine to generate an optimal detection method; receiving a detection result of manually executing the optimal detection method, and inputting the detection result into an inference engine to obtain a posterior probability of the fault tree node; judging whether the posterior probability of a certain fault tree node reaches a preset locking probability or not; if yes, verifying a fault reasoncorresponding to the current fault tree node, pushing a maintenance procedure, and updating the prior probability of the fault tree node according to the current maintenance data; and if not, carrying out a new round of detection method. According to the method provided by the invention, the requirement of the fault tree on the accuracy of artificial experience is reduced, the accuracy of an inference system is improved, the troubleshooting step of fault causes is shortened, and the troubleshooting efficiency is improved.

The invention relates to a data processing device and an artificial intelligence processor, the data processing device is applied to a processing core of the artificial intelligence processor, and thedata processing device is connected with a storage module and a calculation module, and comprises: an address generation module used for generating a first address of first data to be read accordingto a set address generation mode; the data conversion module that is used for reading the first data from the storage module according to the first address and converting the read first data to obtainconverted second data, and sending the second data to a calculation module; and the control module that is used for determining a corresponding address generation mode according to a preset processing category and a data category and controlling the address generation module to generate an address. By generating the address corresponding to the data according to the processing category and the data category and performing data conversion, the access speed of the data can be increased, and the calculation efficiency of the neuromorphic chip is improved.

The invention discloses a reasoning method based on semantic search, comprising: (1) setting a threshold, evaluating the complexity of ABox reasoning and TBox reasoning algorithms, if the algorithms are larger than the threshold, continuing executing, or else, executing step (3); (2) reasoning by the reasoning method, and executing step (4); (3) reasoning by a reasoning method in a Fact++ reasoning machine to get a final reasoning result; (4) modifying a reasoning result got in step (2) to get a final reasoning result. The method can self-adaptively returns a precise reasoning result or an approximate reasoning result, improve reasoning efficiency and lower the time complexity of reasoning.

The invention discloses an object classification method, and the method comprises the steps: constructing an end-cloud combined object classification model in a training process, and the end-cloud combined object classification model comprises an end model and a cloud model; training the end-cloud combined object classification model according to at least one training object; and storing the trained cloud model and the end model. Inference process, the end computing equipment operates the end model to obtain an end model reasoning result; when the confidence coefficient of the end model reasoning result is smaller than or equal to a threshold value, sending the feature vector output by the end model to a cloud model running on cloud computing equipment, and reasoning the feature vector output by the end model sent by the end computing equipment on the cloud computing equipment according to the cloud model to obtain a cloud model reasoning result; and when the confidence coefficient ofthe end model reasoning result is greater than a threshold value, the end model outputs the end model reasoning result. Through the method of combining the end cloud with the object classification, the precision of the object classification is improved.

Techniques for reasoning in a context-aware apparatus are provided. The context-aware apparatus includes least an input, an input selector, and a reasoning component. The information selector selects input information that satisfies a particular condition from among the information received by the input of the context-aware apparatus. The reasoning component performs context-aware reasoning based on the selected input information that satisfies the particular condition.

The invention provides a parallel reasoning algorithm for streaming RDF data, which comprises the steps of building a pseudo bidirectional network of a rule, and building an intermediate node if a class link variable exists in rule nodes; acquiring a batch of new data in a streaming data flow and data generated by previous reasoning at regular time to act as input data, performing classification on the input data or newly building corresponding nodes, and storing the nodes into corresponding Redis clusters; judging whether antecedents monitored by the corresponding intermediate nodes or rule nodes are all satisfied or not by combining the pseudo bidirectional network for the inputted triple data, thus performing reasoning on the rule, and generating reasoning data; and realizing parallel streaming reasoning for an OWL Horst rule of the RDF data integrally and efficiently through deleting repeated reasoning data in real time and storing all data generated by the current time of reasoning into the Redis clusters to act as input data of the next time of reasoning.

The invention relates to a deep learning model reasoning method and device, equipment and a storage medium, and the method comprises the steps: obtaining a to-be-reasoned data set; inputting the data set to be reasoned into a reasoning convolution kernel to obtain a floating point type data reasoning result; obtaining an output scaling factor corresponding to the data set to be reasoned; wherein the output scaling factor is determined according to the maximum value of the elements in the data reasoning result; calculating a product of the data reasoning result and the output scaling factor to obtain a quantification result; and continuing reasoning of the deep learning model according to the quantification result. The method is used for solving the technical problem that in the model performance optimization process in the prior art, the overall error is large due to quantification.

The invention relates to an information processing method and device, computer equipment and a storage medium. The method comprises the following steps: employing an encoder of a machine translation model to encode a to-be-translated word and sentence, and obtaining encoding information; storing the global attention parameter; utilizing a decoder of the machine translation model to determine global attention of the decoder according to the global attention parameter in a decoding cycle of the encoding information of each word in the words and phrases to be translated; and obtaining a prediction result corresponding to a word to be predicted in the decoding loop according to the global attention. Because the global attention parameter required to be used during decoding is stored, the stored global attention parameter can be directly called when the decoder needs to determine the global attention by using the global attention parameter, and the global attention parameter does not need to be obtained by processing the coding information again, so that the calculation amount in the decoding process is reduced. Therefore, the reasoning efficiency of the machine translation model is improved.