59 results about "Probabilistic computing" patented technology

A large language model (LLM) may be used to classify an input into one of a plurality of categories. However, given the machine-learning operation of the LLM, the output of the LLM does not represent a definitive statement, but is based on probability computations of the machine learning model. Therefore, the classification performed by the LLM might not be correct. Classification into the wrong category by the LLM results in downstream technical problems. In some implementations, when an LLM generates a response that classifies an input, one or more probability values associated with a token that forms the basis of the response may be used to determine a confidence value. The confidence value is indicative of confidence in the classification performed by the LLM. An action may be taken based on the confidence value.

The invention provides an automatic data exception root cause positioning and playback repair method and system and related equipment. The method comprises the following steps: constructing a dynamic heterogeneous topological graph containing data logic nodes and physical resource nodes in real time, and establishing real-time directed dependency connection between the nodes; in response to a monitored abnormal signal, generating an anti-fact intervention set assuming that an upstream node is in a reference state by using a Do operator based on the dynamic heterogeneous topological graph; the anti-fact intervention set is substituted into a causal reasoning model for simulation, and the abnormal maintenance probability of the abnormal signal still existing under the condition that the anti-fact intervention set takes effect is obtained; and calculating a causal contribution degree according to the exception maintenance probability to lock a root cause node, and transmitting state data of the root cause node to a repair module to trigger playback repair. According to the invention, the accuracy and stability of abnormal root cause positioning are improved.

A large language model (LLM) may be used to classify an input into one of a plurality of categories. However, given the machine-learning operation of the LLM, the output of the LLM does not represent a definitive statement, but is based on probability computations of the machine learning model. Therefore, the classification performed by the LLM might not be correct. Classification into the wrong category by the LLM results in downstream technical problems. In some implementations, when an LLM generates a response that classifies an input, one or more probability values associated with a token that forms the basis of the response may be used to determine a confidence value. The confidence value is indicative of confidence in the classification performed by the LLM. An action may be taken based on the confidence value.

The invention discloses a fault tree Boolean function equivalent mapping method based on an untrained neural network, and relates to the field of fault tree analysis. In order to solve the problems that in the prior art, a Boolean function mapping structure is not beneficial to parallel expansion, the calculation efficiency is limited, and the Boolean function mapping structure is difficult to efficiently realize on high-parallel platforms such as a GPU, the invention provides a method for generating topological structure data by analyzing a fault tree model; the basic events, the intermediate events and the top events are mapped into neurons of an input layer, a hidden layer and an output layer respectively, a feedforward network with fixed weight and bias is constructed, and a logic activation function is defined in nodes to realize Boolean logic propagation. The input layer receives a basic event state vector, outputs a top event result through forward propagation, and can realize large-scale Boolean function mapping on a parallel platform through batch input matrixes. The method is suitable for reliability analysis, minimum cut set simplification, top event probability calculation, parallelization fault tree solving and the like of a large-scale complex system.

The automatic control method for oil and gas pipeline detection specifically comprises the following steps: step 1, providing basic data; step 2, predicting a future state based on the data; step 3, converting prediction into probability distribution; step 4, calculating a dynamic factor based on probability; by collecting physical attributes of oil and gas pipelines and detecting multi-dimensional state parameters of equipment, comprehensiveness and accuracy of scheduling decision making are achieved; the TCN network based on multi-head attention can effectively capture the time sequence dependency relationship of the pipeline and equipment states, so that the accuracy of state prediction is improved, and a reliable basis is provided for scheduling decision making; the factor graph model realizes probability quantitative analysis of the pipeline state and the equipment state, and can scientifically evaluate the execution risk of the detection task; a time decay factor and a residual electric quantity redundancy value factor are introduced, dynamic changes of task urgency and equipment electric quantity redundancy are considered, and task priority ranking is optimized; through a real-time feedback and dynamic adjustment mechanism, the scheduling scheme can adapt to the state change in the detection process, and the detection efficiency and the resource utilization rate are improved.

The invention relates to the technical field of building engineering construction, in particular to a building engineering construction collaborative management method and system based on the Internet of Things. The method comprises the following steps: acquiring multi-modal data of a construction site through Internet of Things equipment; constructing a construction knowledge graph; selecting a small amount of data marked with construction risks from multi-modal data as a sample data set to train an initial risk prediction model, generating a cross-modal consistency constraint extension training set through pseudo labels, and dynamically updating the risk prediction model based on an incremental learning method of a sliding window; inputting the construction knowledge graph into the heterogeneous graph neural network, and propagating and aggregating the characteristics of nodes and edges; identifying a risk causal chain in the construction knowledge graph, performing virtual intervention based on a causal inference method, calculating the intervention probability of subsequent node risks, and calculating the causal priority of each risk event according to the intervention probability; and retraining the risk prediction model according to feedback of the management platform.

The invention discloses a maintenance strategy optimization and task assignment method for new energy equipment, and the method comprises the steps: collecting multi-source operation data of a new energy station, and carrying out the preprocessing of the multi-source operation data, so as to form an operation state data set; establishing an operation life and performance attenuation model to predict a fault probability, and calculating a power generation loss risk, a safety risk and a chain shutdown risk of each device of the station; identifying a maintenance opportunity window, and constructing a risk-opportunity joint index; aggregating the maintenance objects into a task package according to geographical proximity, an isolation relationship and an operation type, and determining a task step chain in the task package according to operation dependence and process logic; in a rolling time domain, an execution sequence and a time interval of a task package are optimized, a maintenance task plan and a resource scheduling scheme are generated, risk analysis, opportunity recognition, task aggregation, optimization scheduling and dynamic feedback are integrated, and intelligent decision making, real-time optimization and system collaboration of a new energy station maintenance process are realized.

The invention discloses an elevator prediction scheduling method and system based on user habit self-learning, and the method comprises the steps: carrying out the statistical analysis based on historical elevator taking event data, and judging whether a floor combination meeting a preset condition exists or not; combining all the independent floors with the floors meeting the preset conditions to serve as a self-learning unit; for each self-learning unit, updating the elevator calling probability of the respective learning unit by using an exponential weighted average algorithm; when the elevator does not have the real-time task, all elevator calling probabilities corresponding to the current time window are inquired, and the self-learning units meeting the triggering condition are extracted as a target candidate set; calculating an optimal pre-stop layer by utilizing a mathematical model based on the target candidate set, and calculating a comprehensive probability corresponding to the optimal pre-stop layer; and calculating an income evaluation function based on the comprehensive probability, and generating an instruction to drive the elevator to run to an optimal pre-stop layer to enter a prediction waiting state when an obtained value meets a condition.

This invention relates to the field of artificial intelligence technology, providing a text segmentation method, system, computer device, and storage medium, comprising: acquiring the text to be segmented; dividing each sentence of the text into multiple semantic blocks according to a left-to-right order, with the end point as the boundary; if two consecutive characters do not have a connected record in the vocabulary of the text to be segmented, then the preceding character of the two consecutive characters is recorded as the end point; performing full segmentation on each semantic block to obtain all possible segmentation methods for that semantic block; calculating the probability of traversing all segmentation methods for each semantic block in a left-to-right order, and selecting the segmentation method with the highest probability as the final segmentation result. The segmentation scheme of this invention is based on probability, traverses all solutions within a block, and comprehensively considers the context of the text, resulting in more accurate segmentation results, reducing labor costs, and improving segmentation accuracy.

The invention belongs to the technical field of artificial intelligence, and provides a task planning method and device based on large model feedback optimization, and the method comprises the steps: converting the natural language description of a problem into a PDDL symbol sequence, and calculating an output probability corresponding to the PDDL symbol sequence according to an LoRA parameter; calculating a loss value between the output probability and the target PDDL symbol sequence so as to update LoRA parameters, and training to obtain a large language model with a planning generation capability; and performing executable verification on the candidate PDDL planning scheme output by the model, adjusting input context information input into the model by using an error log, and obtaining an executable planning scheme through an iterative verification process. According to the method and the device, new errors caused by full-amount rewriting are avoided by limiting the modification range, and solvability and verification pass of a planner are taken as a stop condition during iteration of the model, so that a result has evaluability and reproducibility.

The invention relates to a power network lightweight attack path generation method and system, and belongs to the technical field of attack path generation, and the method comprises the steps: obtaining the power network flow data of a target regional power grid in real time, carrying out the preprocessing, eliminating the sampling difference between heterogeneous devices, and obtaining the preprocessed time sequence aligned heterogeneous data; carrying out heterogeneous graph modeling based on the preprocessed time sequence aligned heterogeneous data to obtain a combined heterogeneous graph; wherein the heterogeneous graph modeling comprises the steps of extracting differentiated node features based on a source equipment type, and constructing edges with interaction features based on power business logic and a protocol type to form a protocol interaction element path; inputting the combined heterogeneous graph into a pre-trained potential attack path generation model to obtain a potential attack path probability; wherein the potential attack path generation model is obtained by performing joint adversarial training on the potential attack path generation model and the noise generation model; a suspicious path score is calculated based on the potential attack path probability, and an attack path is generated based on the suspicious path score.

This invention provides a spatially parallel hybrid multiplier based on probabilistic computation and its operating method. The spatially parallel hybrid multiplier based on probabilistic computation includes a traditional multiplier and two high-precision probabilistic multipliers. Each high-precision probabilistic multiplier includes a random sequence generator, a random computation circuit, a probability estimator, and a two's complement conversion circuit connected in sequence. The spatially parallel hybrid multiplier based on probabilistic computation provided by this invention successfully utilizes the characteristics of probabilistic computation, implementing multiplication operations using only a few wires, logic gates, and single-bit addition, thus reducing computational complexity and resource overhead. Simultaneously, the spatially parallel hybrid multiplier based on probabilistic computation provided by this invention strikes a good balance between computational accuracy and hardware resources, ensuring computational accuracy while significantly reducing hardware resources. This saves area and reduces power consumption.

A large language model (LLM) may be used to classify an input into one of a plurality of categories. However, given the machine-learning operation of the LLM, the output of the LLM does not represent a definitive statement, but is based on probability computations of the machine learning model. Therefore, the classification performed by the LLM might not be correct. Classification into the wrong category by the LLM results in downstream technical problems. In some implementations, when an LLM generates a response that classifies an input, one or more probability values associated with a token that forms the basis of the response may be used to determine a confidence value. The confidence value is indicative of confidence in the classification performed by the LLM. An action may be taken based on the confidence value.

The invention provides a probabilistic nearest neighbor query method on high-dimensional continuous uncertain data, and relates to a method for quickly retrieving the nearest neighbor of the uncertain data by combining a locality sensitive hashing technology and probability calculation. The method comprises the following steps: firstly, constructing locality sensitive Hash for mapping, and mapping the uncertainty of a data object into a plurality of Hash tables through sampling; in a query stage, candidate adjacent objects are extracted from a given query point by using an index, and the probability that each candidate object becomes the nearest neighbor of the query point is calculated through Monte Carlo simulation. And finally outputting the neighbor object with the maximum probability and the probability value thereof. According to the method, the efficiency of high-dimensional uncertain data nearest neighbor query is greatly improved while the query accuracy is guaranteed, and the method can be widely applied to the fields of big data analysis, uncertain information retrieval and the like.

This application discloses an automatic control method for oil and gas pipeline inspection, specifically comprising the following steps: Step 1 provides basic data; Step 2 predicts future states based on the data; Step 3 transforms the predictions into probability distributions; Step 4 calculates dynamic factors based on probability; Step 5 achieves optimal scheduling using probability and factors; and Step 6 maintains scheduling timeliness through real-time feedback. By collecting the physical properties of oil and gas pipelines and multi-dimensional state parameters of inspection equipment, comprehensive and accurate scheduling decisions are achieved. A multi-head attention-based TCN network effectively captures the temporal dependencies between pipeline and equipment states, improving the accuracy of state prediction and providing a reliable basis for scheduling decisions. A factor graph model enables probabilistic quantitative analysis of pipeline and equipment states, scientifically assessing the execution risks of inspection tasks. The introduction of time decay factors and remaining power redundancy value factors considers the dynamic changes in task urgency and equipment power redundancy, optimizing task priority ranking. Through real-time feedback and dynamic adjustment mechanisms, the scheduling scheme adapts to state changes during the inspection process, improving inspection efficiency and resource utilization.

The application discloses a time sequence early classification method, a terminal device and a storage medium, constructs a training set by using time sequence data of human body actions; trains a neural network by using the training set; inputs the training set into the trained neural network to obtain classification probabilities of all moments of all data of the training set, calculates a probability exit threshold value by using the classification probabilities; inputs observable data of a moment t into the trained neural network to obtain a classification probability Pt of the moment t, takes a maximum value of the Pt, and if the maximum value is greater than the probability exit threshold value, stops inputting the observable data, and takes a classification result of the moment t as a final classification result of the observable data of the moment t. The application can adapt to continuously increasing new data, extract more distinguishing features, and improve the accuracy of early classification of time data; the application can adapt to sample content and difficulty, extract more class-specific features, and improve the accuracy of early classification.

A large language model (LLM) may be used to classify an input into one of a plurality of categories. However, given the machine-learning operation of the LLM, the output of the LLM does not represent a definitive statement, but is based on probability computations of the machine learning model. Therefore, the classification performed by the LLM might not be correct. Classification into the wrong category by the LLM results in downstream technical problems. In some implementations, when an LLM generates a response that classifies an input, one or more probability values associated with a token that forms the basis of the response may be used to determine a confidence value. The confidence value is indicative of confidence in the classification performed by the LLM. An action may be taken based on the confidence value.

The invention discloses a probabilistic power flow algorithm based on a linear power flow model and a total probability formula, and aims at a scene that a conventional probabilistic power flow calculation method based on semi-invariant cannot be carried out due to the ill-conditioned Jacobian matrix in probabilistic power flow analysis of an active power distribution network. And an optimal state variable function is selected to improve the calculation precision. A total probability formula is introduced in probability calculation, N output interval combinations are obtained by performing segmentation processing on the new energy unit, and a joint conditional probability density function and each-order semi-invariant of the new energy unit under the N output interval combinations are calculated. The method comprises the following steps of: calculating a semi-invariant of a state variable based on a linear power flow model under each output interval combination, finally fitting a distribution function by adopting a Gram-Charlier (GC) series to obtain N conditional probability distribution functions of the state variable, and aggregating by adopting a total probability formula to obtain a complete probability density function. According to the method, the precision of probabilistic power flow analysis is improved, and the technical method has relatively high implementability.

The invention provides a model reasoning method and device and related equipment, and relates to the technical field of artificial intelligence, the model reasoning method comprises the steps that a target cue word is sent to second electronic equipment, the target cue word is used for enabling a reasoning model deployed on the second electronic equipment to conduct reasoning according to the target cue word, and the target cue word is sent to the second electronic equipment; generating a first reasoning result; the first reasoning result sent by the second electronic equipment is received, the first reasoning result is input into a verification model for probability calculation, a target probability is obtained, the target probability is used for representing the probability that the first reasoning result is the reasoning result corresponding to the target prompt word, and the target prompt word is sent to the second electronic equipment. The reasoning model is determined according to the verification model; and under the condition that the target probability is greater than an acceptable probability, determining the first reasoning result as a reasoning result corresponding to the target cue word. Therefore, the output efficiency of the reasoning result corresponding to the target prompt word can be improved.

A computer-implemented system and method for determining collapse states in quantum and probabilistic systems using wave-interference-based computation are disclosed. The system models collapse as a deterministic process arising from interaction between a system wavefunction and one or more observer or environmental wavefunctions. A modified Schrödinger-type formulation is used to compute a real-valued collapse intensity based on amplitude coupling and phase alignment between interacting wavefunctions. The system further aggregates the collapse intensity over a spatial domain to generate a scalar collapse measure, which is compared against a predefined threshold to determine collapse conditions. The invention enables simulation, prediction, and control of collapse behavior across quantum systems, probabilistic computation, and collapse-driven decision architectures.

The application discloses a risk assessment method and system based on fault semantic driving, a medium and a product, and relates to the field of power new energy equipment detection. According to the method, after a power new energy equipment detection system obtains a power equipment fault description text input by an operation and maintenance personnel, the power new energy equipment detection system matches the fault description text with fault semantic elements corresponding to nodes in a fault evolution track graph, and calculates the current risk intensity of the target node. At the same time, the power new energy equipment detection system calculates the current risk transmission value of the target node through a target directed edge to connect a downstream node according to an initiation probability corresponding to the target directed edge, and determines a current fault risk vector of the power equipment at a current time by combining an initial fault risk vector of the power equipment at an initial time, so as to determine a current risk assessment value of the power equipment. The fault diagnosis method based on risk propagation and accumulation can dynamically track and evaluate the fault diffusion process, and improves the accuracy and timeliness of fault detection.

The application relates to the technical field of data processing, in particular to a video intention understanding method and system based on implicit behavior entropy. The method comprises the following steps: for any video segment, collecting real-time interactive behaviors of a user on the video segment to obtain an implicit interactive sequence; calculating implicit behavior entropy of the video segment according to occurrence probabilities of various interactive behaviors in the implicit interactive sequence; dynamically updating weights of edges between video segment nodes and creative concept nodes based on the implicit behavior entropy; for any video segment, calculating semantic association abundance according to the weights of the edges connected with the video segment, and calculating a creative value score of the video segment according to the semantic association abundance of the video segment. The method can deeply mine the value of the video and improve the experience of the creator and the user.

The application provides an automatic analysis method and device for key influencing factors of digitalized power grids, and relates to the technical field of power grid data processing. The method comprises the following steps: issuing a domain language model and a variational topic model to multiple power grid nodes; receiving local topic word probabilities obtained by processing private corpus of each power grid node by the domain language model and the variational topic model; calculating global topic word probabilities according to the local topic word probabilities of the multiple power grid nodes; calculating comprehensive weights according to the global topic word probabilities and frequencies of occurrence of each topic word; taking a topic word with a comprehensive weight greater than a preset weight value as an influencing factor, and obtaining index features and technical features associated with the influencing factor; and constructing a multi-layer influencing factor architecture according to the influencing factor, the index features and the technical features, so as to reflect fluctuations in contribution degrees of each influencing factor to the digitalized power grid over time, thereby achieving the purpose of quantifiable mining of key influencing factors.

The application provides a code generation method and device based on token appearance probability adjustment, and a computer device. The method comprises the following steps: screening a plurality of candidate tokens satisfying constraints from a preset token table; calculating an adjusted appearance probability of a second target token according to a plurality of original appearance probabilities corresponding to the plurality of candidate tokens and an original appearance probability of the second target token; adjusting the original appearance probability of the second target token to the adjusted appearance probability; performing normalization adjustment on the original appearance probabilities of the plurality of candidate tokens to obtain adjusted appearance probabilities of the plurality of candidate tokens; calculating appearance probabilities of a plurality of second candidate codes according to the adjusted appearance probabilities of the plurality of candidate tokens and the adjusted appearance probability of the second target token; and generating a target code according to the appearance probabilities of the plurality of second candidate codes. The application can ensure that the probability distribution of the language model is not distorted, and the code generation accuracy of the language model is improved.

The invention discloses a sample injection method and device of a model, equipment and a storage medium, and relates to the technical field of computers, the method comprises the following steps: firstly, obtaining a business scene corresponding to the model and a to-be-injected sample containing text, structured and image data; extracting sample features, coding a service scene to obtain context features, and calculating an injection weight according to the context features; inputting a sample into the model, determining sample loss in combination with a real label to calculate a parameter mask, and controlling an injection process through the two; extracting a sample key part, determining a risk score according to a prediction probability average value of the sample key part in an irrelevant scene, and adjusting an injection weight and a parameter mask when the risk score is smaller than a risk threshold value; and finally, obtaining an output result and a target output probability before and after sample injection, calculating output consistency and a capability drift score, and when the output consistency score is lower than a first threshold value or the capability drift score exceeds a second threshold value, recovering the model to the previous version.

This disclosure relates generally to a ferroelectric field effect transistor (FeFET) device suitable for probabilistic computing. The FeFET devices is usable as tunable probabilistic element in a network. A computing device comprising the network is also proposed. The FeFET device comprising a channel layer comprising an oxide semiconductor material and / or a 2D semiconductor material. The FeFET device further comprises a gate structure on the channel layer, the gate structure comprising a ferroelectric or anti-ferroelectric layer and a gate layer arranged on the ferroelectric or anti-ferroelectric layer. The gate structure is arranged between a source contact and a drain contact, which are additionally arranged on the channel layer.

The present invention implements a probabilistic bit element with a simple circuit structure consisting of a threshold switch element and a resistor. The threshold switch element comprises: a lower electrode; a stacked structure provided above the lower electrode; and an upper electrode provided above the stacked structure, wherein the stacked structure includes a silicon oxide layer and an oxygen vacancy supply layer for supplying oxygen vacancies to the silicon oxide layer when an input voltage is applied, and oscillation of an output voltage can be stochastically generated.

The invention discloses an industrial network intelligent decision-making system and method based on digital twinning. The method comprises the following steps: collecting multi-source sensor data in an industrial field, and forming a standard data vector through wavelet threshold denoising and long and short-term memory network missing value restoration; the method comprises the following steps: constructing geometric, physical and process models of a production line in a digital twin server, and realizing time and state synchronization through an IEEE 1588 PTP protocol; inputting the standard data vector into an Informer-based time sequence prediction model, and realizing fault type and probability calculation and fault position identification in combination with equipment dimension coding; according to a detection result, a multi-objective optimization decision model is established by using a genetic algorithm, and an optimal maintenance and scheduling scheme is generated by integrating shutdown loss, maintenance cost and production guarantee; and issuing the scheme to a local controller for execution. The system realizes a closed-loop decision flow from data acquisition, modeling and detection to optimization and feedback.

The application belongs to the field of magnetic devices, and relates to a magnonic random number generator, a magnonic probability calculation integrated device and a probability generation method. The magnonic random number generator comprises a pulse input module, a random number generation module and an output module, and the input end of the random number generation module and the output end of the random number generation module are connected with the pulse input module and the output module through an adhesive layer. The random number generation module is used for generating random pulses with the bistable characteristic of strong nonlinear effect under the condition of receiving pulse signals from the pulse input module and receiving the random pulses by the output module. The random number generation module is a magnetic waveguide or a thin film with a high nonlinear frequency shift coefficient and low damping deposited on an insulating substrate. The random number generated by the magnonic random number generator can reduce energy consumption, improve the generation rate and thermal stability of the random number, and can be further used for probability calculation of the random number.