A Dynamic Analysis Method for Event Clues Based on Large Model Intelligent Agents

Abstract

This invention discloses a dynamic event clue analysis method based on a large-scale intelligent agent, comprising: acquiring multimodal event clues generated by the large-scale intelligent agent and forming a unified structured data input; extracting key semantic feature vectors of the event clues and encoding them as initial quantum state representations; using parameterized quantum gates to perform quantum entanglement evolution to calculate the attention weight matrix; performing quantum feature field modeling and defining dynamic clue potential energy functions; determining clue convergence points and event causal paths to construct an event association graph; evaluating the causal correlation degree of events in real time through quantum probabilistic reasoning to obtain the event evolution path; adaptively adjusting the attention weight matrix based on semantic feature similarity and iteratively calculating until convergence, outputting the final event clue evolution path. This invention improves the accuracy and real-time performance of dynamic association analysis of heterogeneous event clues.

Description

Technical Field

[0001] This invention relates to the field of multimodal data analysis technology, and in particular to a method for dynamic analysis of event clues based on large model intelligent agents. Background Technology

[0002] With the rapid development of artificial intelligence (AI) technology, multimodal information analysis technology based on large-scale AI models has become increasingly mature and widely applied in the field of event monitoring and analysis. Currently, multimodal event cues generated by large-scale AI models typically include text, images, voice, and various sensor data. Existing technologies usually first convert event cues of different modalities into structured data, and then use semantic feature extraction, fusion, and correlation analysis methods to achieve causal relationship reasoning and evolution path prediction between events, in order to meet the needs of event monitoring and analysis in complex scenarios.

[0003] Currently, the main methods for correlation analysis of multimodal event cues include attention mechanisms based on deep learning networks and feature fusion analysis methods based on graph neural network structures. Attention mechanisms quantify the influence of different event cues on the target event by calculating semantic similarity weights between event cues; graph neural networks utilize graph structures to model the correlations between event cues to achieve fusion analysis and prediction of event evolution paths. However, existing attention mechanisms and graph neural networks often suffer from low feature fusion efficiency, unstable attention weight calculation, and insufficient accuracy in dynamic correlation analysis of event cues when processing multimodal event cues generated by large-scale artificial intelligence models. They struggle to accurately identify implicit nonlinear correlation features and potential causal relationships evolving over time. Since existing technologies primarily employ classical neural networks and traditional probabilistic reasoning methods for feature fusion and event evolution path prediction, their computational efficiency is insufficient for real-time processing of massive heterogeneous event cues, and they cannot meet the real-time and accuracy requirements of highly dynamic and ever-changing event analysis scenarios.

[0004] Therefore, how to provide a dynamic analysis method for event clues based on large-scale intelligent agents is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a dynamic judgment method for event clues based on large-scale intelligent agents. Addressing the problem of low accuracy in real-time and precise correlation analysis of heterogeneous multimodal event clues in existing technologies, this invention proposes a dynamic judgment method based on quantum adaptive attention deep fusion quantum neural field network. Through quantum feature field modeling, quantum probabilistic reasoning, and adaptive adjustment of semantic feature similarity, the accuracy and efficiency of real-time correlation analysis of heterogeneous event clues are improved.

[0006] The event clue dynamic analysis method based on a large model intelligent agent according to an embodiment of the present invention includes:

[0007] Acquire multimodal event cues dynamically generated by large model agents and construct structured data inputs;

[0008] For each type of event clue in the structured data input, corresponding feature extraction is performed to obtain key semantic feature vectors, which are then encoded into the corresponding initial quantum state representation;

[0009] The initial quantum state representation is processed using pre-set parameterized quantum gates to obtain the quantum attention weight matrix;

[0010] Continuous quantum characteristic field modeling is performed for all event clues, and a differentiable dynamic clue potential energy function is defined;

[0011] Calculate the local extrema of the dynamic cue potential energy function in the continuous quantum characteristic field and the corresponding potential energy gradient path, determine the potential cue convergence point and event causal path of event cues, and construct the event correlation graph;

[0012] Based on the event association diagram, historical clue paths, and newly added event clues within the current time window, confidence assessment is performed to obtain the event evolution path;

[0013] The event clues corresponding to each event node in the event evolution path are mapped back to the key semantic feature vectors. The quantum attention weight matrix is ​​adjusted according to the semantic feature similarity calculation results. The continuous quantum feature field reconstruction and event causal path calculation are iteratively executed to obtain the final event clue evolution path.

[0014] Optionally, encoding the key semantic feature vector into the corresponding initial quantum state representation specifically involves:

[0015] Perform multi-dimensional feature normalization on the key semantic feature vectors of each event clue to generate normalized key semantic feature vectors;

[0016] The number of qubits required in the corresponding initial quantum state representation is determined based on the number of feature dimensions contained in the normalized key semantic feature vector, and an amplitude range is preset for each qubit, and the amplitude range is divided into multiple non-overlapping sub-intervals.

[0017] Based on the pre-defined correspondence between the feature dimensions and the sub-intervals, an initial quantum bit amplitude distribution is formed;

[0018] A second normalization operation is performed on the initial qubit amplitude distribution to obtain the quantum state amplitude distribution;

[0019] Based on the quantum state amplitude distribution, an initial phase parameter is set for each qubit in the initial quantum state representation;

[0020] According to the preset quantum bit arrangement order, the quantum state amplitude distribution and initial phase parameters are loaded one by one into the initial quantum state representation to obtain the initial quantum state representation corresponding to the key semantic feature vector.

[0021] Optionally, the process of processing the initial quantum state representation using pre-set parameterized quantum gates to obtain the quantum attention weight matrix is ​​as follows:

[0022] Based on the amplitude and phase values ​​of each qubit in the initial quantum state representation, determine the sorting sequence of amplitude values ​​from high to low and the sorting sequence of phase values ​​from small to large for all qubits in the initial quantum state of each event clue;

[0023] All qubits in the initial quantum state representation of each event clue are rearranged one by one in an alternating order of amplitude and phase to form an adaptive qubit recombination sequence;

[0024] The adaptive qubit recombination sequence is input one by one into a parameterized quantum gate to perform the initial entanglement evolution;

[0025] Perform a phased probability measurement operation based on the quantum state after the initial entanglement evolution, determine the probability distribution of the current quantum state through the phased probability measurement results, and record the probability value of each quantum bit after measurement at the same time.

[0026] Based on the phase difference between the phase measurement results and the phase difference between each quantum bit in the current quantum state, the secondary sorting sequence of the quantum bits is determined, and secondary quantum bit recombination is performed.

[0027] Repeatedly execute quantum state sequence input, real-time adaptive determination of rotation gate rotation parameters, staged probability measurement and adaptive recombination of qubits to complete multiple rounds of entanglement evolution;

[0028] After completing all rounds of entanglement evolution, the joint probability distribution matrix of all qubits is obtained by performing joint projection measurements on the global basis of quantum states.

[0029] The quantum state correlation strength between the initial quantum states of any two event cues is calculated based on the joint probability distribution matrix, and then filled into the quantum attention weight matrix in the order of the event cues to obtain the complete quantum attention weight matrix.

[0030] Optionally, the pre-set parameterized quantum gate is specifically:

[0031] A quantum gate sequence consisting of alternating series of rotating gates and controlled NOT gates, wherein the initial values ​​of the rotation axis direction and rotation angle parameters of the rotating gate are determined based on the centroid and standard deviation of the amplitude distribution of all qubits in the input qubit recombination sequence;

[0032] The pairing order of the control qubit and the target qubit of the controlled NOT gate is determined by sorting the amplitude differences between adjacent qubits in the input qubit recombination sequence from largest to smallest;

[0033] In each round of entanglement evolution, the rotation axis direction and rotation angle parameters of the rotating gate are updated in real time based on the stage probability measurement values ​​obtained from the previous round of evolution. The pairing order of the control qubit and the target qubit of the controlled NOT gate is adjusted and updated based on the qubit amplitude change determined by the stage probability measurement values ​​of the previous round.

[0034] Optionally, the continuous quantum characteristic field modeling of all event clues and the definition of a differentiable dynamic clue potential function are as follows:

[0035] The sum of the overall attention weights for each event cue is determined based on the quantum attention weight matrix, and used as the initial value for the spatial density;

[0036] Using the amplitude and phase values ​​of key semantic feature vectors as the spatial coordinate axes of the continuous quantum feature field, the spatial location coordinates of event clues are determined based on the initial value of spatial density.

[0037] Construct joint coordinate points for event clues based on their spatial location coordinates and time tags;

[0038] A cubic spline interpolation method is used to perform continuous smooth mapping of the joint coordinate points of the event clues, thereby obtaining the complete spatiotemporal continuous distribution of the event clues;

[0039] Determine the local spatial density function based on the complete spatiotemporal continuous distribution;

[0040] A dynamic cue potential energy function is constructed based on the local spatial density function. The function value of the dynamic cue potential energy function at any spatial location is the negative logarithm of the local spatial density function value at the corresponding location, and it satisfies the continuous differentiability property.

[0041] Optionally, the calculation of the local extrema of the dynamic clue potential energy function in the continuous quantum characteristic field and the corresponding potential energy gradient path, determining the potential clue convergence point and event causal path, and constructing an event correlation graph, specifically involves:

[0042] Based on the dynamic clue potential energy function, the spatial coordinates of the local extremum points are solved using the multidimensional Newton iteration method;

[0043] Tracing the gradient path along the negative gradient direction of the dynamic cue potential energy function to determine potential cue convergence points;

[0044] The starting and ending nodes of the event causal path are determined based on the starting and ending positions of the gradient path.

[0045] The starting and ending nodes of the event causal path are used as event nodes in the event association graph, and directed association edges are defined between event nodes.

[0046] The temporal sequence of associated edges is determined by the time tag relationship between event nodes, and an event association graph with temporal relationship is constructed.

[0047] Optionally, the step of performing confidence assessment based on the event association graph, historical clue paths, and newly added event clues within the current time window to obtain the event evolution path specifically involves:

[0048] Based on the attention weight values ​​of all event clues in the historical clue path and the attention weight values ​​of newly added event clues, the cumulative value of the attention weights of all connected event clues is calculated to form the initial quantum probability amplitude distribution.

[0049] The amplitude values ​​of all event nodes in the initial quantum probability amplitude distribution of each event node are mapped one by one to the qubits of the corresponding event nodes in the quantum probability inference space, and the amplitude value of each qubit is determined to generate the initial quantum probability state.

[0050] Construct a quantum operator matrix based on the directed association edges between any two event nodes in the event association graph;

[0051] The constructed quantum operator matrix is ​​applied to the initial quantum probability state in the quantum probability inference space to obtain a set of joint quantum probability amplitude values.

[0052] Perform an amplitude squaring operation on the joint quantum probability amplitude value of each event causal path to obtain a set of real-time probability values;

[0053] The real-time probability values ​​of all event causal paths in the real-time probability value set are compared and sorted one by one, and the event causal path with the highest real-time probability value is selected as the event evolution path within the current time window.

[0054] Optionally, the step of mapping the event clues corresponding to each event node in the event evolution path back to the key semantic feature vector, adjusting the quantum attention weight matrix according to the semantic feature similarity calculation result, and iteratively performing continuous quantum feature field reconstruction and event causal path calculation to obtain the final event clue evolution path is as follows:

[0055] Map each event node in the event evolution path obtained in the current time window back to the event clue corresponding to the event node in the structured data input, and obtain the original key semantic feature vector of the event clue corresponding to each event node;

[0056] Based on the key semantic feature vectors of the event nodes, calculate the cosine similarity of the corresponding key semantic feature vectors for any two event nodes that are connected by a directed edge.

[0057] The numerical difference between the cosine similarity between each pair of related event nodes and the attention weight values ​​between the corresponding event nodes in the quantum attention weight matrix is ​​calculated to obtain a set of numerical differences.

[0058] The numerical differences in the set of numerical differences are summed to obtain the total numerical difference of the current quantum attention weight matrix, and the total numerical difference is compared with a preset convergence threshold.

[0059] When the cumulative value of the numerical difference exceeds the preset convergence threshold, the attention weight values ​​of the quantum attention weight matrix are updated by scaling the attention weight values ​​between the corresponding event nodes in the quantum attention weight matrix.

[0060] Based on the updated quantum attention weight matrix, the continuous quantum feature field modeling, dynamic cue potential function, potential gradient path calculation, event association graph reconstruction, and quantum probabilistic inference of event evolution path are re-executed to obtain the updated event evolution path.

[0061] The beneficial effects of this invention are:

[0062] (1) This invention achieves efficient and accurate correlation analysis between heterogeneous multimodal event clues by integrating quantum feature field modeling and quantum probabilistic reasoning based on quantum adaptive attention mechanism, effectively improving the accuracy and real-time performance of dynamic correlation of event clues, and enhancing the reliability of event evolution path identification and prediction in complex scenarios.

[0063] (2) This invention achieves effective identification and reasoning of potential causal relationships between event clues through gradient path analysis of dynamic clue potential energy function and event association graph construction method, significantly improves the reasoning accuracy of event evolution path, and shows better adaptability and stability in multimodal heterogeneous data scenarios.

[0064] (3) In the real-time analysis of multimodal event clues, this invention effectively solves the problems of low feature fusion efficiency and unstable attention weight calculation in the prior art by adaptively adjusting the quantum attention weight matrix through semantic feature similarity. It breaks through the bottleneck of the prior art in the identification of nonlinear event association features, achieves a significant improvement in the accuracy of event clue association, and effectively enhances the practical application level of dynamic judgment of multimodal event clues in the field of artificial intelligence. Attached Figure Description

[0065] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0066] Figure 1 This is a flowchart of the event clue dynamic judgment method based on large model intelligent agents proposed in this invention. Detailed Implementation

[0067] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0068] refer to Figure 1 A dynamic event clue analysis method based on large-scale intelligent agents includes:

[0069] It acquires multimodal event cues dynamically generated by large model agents, including text, images, voice, and sensor data, and parses the raw data content and corresponding time tags of each event cue to form a unified structured data input.

[0070] For each type of event clue in the structured data input, the corresponding feature extraction is performed to obtain the key semantic feature vector of each event clue in the original feature space, and the key semantic feature vector is encoded into the corresponding initial quantum state representation;

[0071] By using pre-set parameterized quantum gates to perform multiple rounds of iterative entanglement evolution and quantum state measurement operations on the initial quantum state representation, the attention weight values ​​between any two event cues are calculated, and the complete quantum attention weight matrix between event cues is obtained.

[0072] Based on the attention weights of each event cue in the quantum attention weight matrix and the time labels of the corresponding event cue, a continuous quantum feature field model is performed for all event cue, and a differentiable dynamic cue potential energy function is defined based on the spatial distribution of event cue in the quantum feature field.

[0073] Calculate the local extrema of the dynamic clue potential energy function in the continuous quantum characteristic field and the corresponding potential energy gradient path, determine the potential clue convergence point and event causal path of the event clue, and construct the event association graph based on the potential clue convergence point and event causal path;

[0074] Based on the event association diagram, historical clue paths, and newly added event clues within the current time window, the quantum probabilistic reasoning method is used to perform real-time confidence assessment of the correlation and causal relationship between events, and to obtain the event evolution path with the highest causal correlation.

[0075] The event clues corresponding to each event node in the event evolution path are mapped and compared with the key semantic feature vectors in the original feature space. The semantic feature similarity between the key semantic feature vectors is calculated, and the quantum attention weight matrix is ​​adaptively adjusted according to the semantic feature similarity. The continuous quantum feature field reconstruction and event causal path calculation are iteratively executed until the similarity between the quantum attention weight matrix and the semantic feature in the original feature space simultaneously meets the preset convergence condition, and the final event clue evolution path is output.

[0076] In this embodiment, the acquisition of multimodal event cues dynamically generated by the large model agent, including text, images, voice, and sensor data, and the parsing of the original data content and corresponding time tags of each event cue to form a unified structured data input, specifically involves:

[0077] Collect multimodal event cue data generated in real time by a large model intelligent agent. The multimodal event cue data includes text data, image data, voice data, and sensor data.

[0078] Perform semantic parsing on the collected text data to extract semantic entities, keywords, and semantic structures that reflect event attributes from the text data, and record the timestamp of the text data generation;

[0079] Perform feature segmentation and target recognition operations on the acquired image data to determine the object category, appearance features and location coordinates associated with the event in the image data, and record the timestamp of the image data generation;

[0080] The system performs speech feature extraction and speech recognition on the collected speech data to determine the text transcription content, speech emotion features, and speaker identity features corresponding to the speech data, and records the timestamp of the speech data generation.

[0081] Perform signal preprocessing and feature extraction operations on the collected sensor data to determine the corresponding physical measurement parameters and state characteristics in the sensor data, and record the timestamp of the sensor data generation.

[0082] The features in the parsed text data, image data, voice data, and sensor data are structured and encoded with their corresponding timestamps to form a unified structured data input.

[0083] In this embodiment, encoding the key semantic feature vector into the corresponding initial quantum state representation specifically involves:

[0084] Perform multi-dimensional feature normalization processing on the key semantic feature vector of each event clue, constrain the value of the feature component of each dimension in the key semantic feature vector to between 0 and 1, and generate a normalized key semantic feature vector.

[0085] The number of qubits required in the corresponding initial quantum state representation is determined based on the number of feature dimensions contained in the normalized key semantic feature vector. An amplitude range is preset for each qubit, and the amplitude range is divided into multiple non-overlapping sub-intervals, so that each sub-interval uniquely corresponds to a feature dimension.

[0086] Based on the pre-defined correspondence between feature dimensions and sub-intervals, the values ​​of each feature dimension in the normalized key semantic feature vector are mapped one by one to the sub-intervals of their respective qubit amplitudes, forming the initial qubit amplitude distribution.

[0087] A quantum state amplitude normalization method is used to perform a second normalization operation on the initial quantum bit amplitude distribution, so that the sum of the squares of all quantum bit amplitude components is strictly equal to 1, thereby obtaining a quantum state amplitude distribution that satisfies the quantum state definition conditions. The quantum state amplitude normalization method is to accumulate the square values ​​of each amplitude component in the initial quantum bit amplitude distribution to obtain the magnitude sum of squares of the amplitude distribution, and use the square root of the magnitude sum of squares of the amplitude distribution as a normalization factor to divide each amplitude component in the initial quantum bit amplitude distribution by the normalization factor, thereby obtaining a quantum state amplitude distribution that satisfies the quantum state definition conditions.

[0088] Based on the quantum state amplitude distribution, an initial phase parameter is set for each quantum bit in the initial quantum state representation. The value of the initial phase parameter is between 0 and 2π, and there is a one-to-one mapping relationship between the setting of the initial phase parameter and the amplitude value of the corresponding feature dimension.

[0089] According to the preset quantum bit arrangement order, the quantum state amplitude distribution and initial phase parameters are loaded one by one into the initial quantum state representation, so that each quantum bit in the initial quantum state representation has both amplitude and phase values, thereby obtaining the initial quantum state representation corresponding to the key semantic feature vector.

[0090] In this embodiment, the step of performing multiple rounds of iterative entanglement evolution and quantum state measurement operations on the initial quantum state representation using pre-set parameterized quantum gates, and calculating the attention weight values ​​between any two event cues to obtain the complete quantum attention weight matrix between event cues, is as follows:

[0091] Before performing multiple rounds of iterative entanglement evolution, based on the amplitude and phase values ​​of each qubit in the initial quantum state representation, determine the sorting sequence of amplitude values ​​from high to low and the sorting sequence of phase values ​​from small to large for all qubits in the initial quantum state of each event clue.

[0092] All qubits in the initial quantum state representation of each event clue are rearranged one by one in an alternating order of amplitude and phase to form an adaptive qubit recombination sequence that differs from the initial quantum state representation and has a definite sequence.

[0093] The adaptive qubit recombination sequence is input one by one into a pre-set parameterized quantum gate for the initial entanglement evolution, so that the quantum state after the initial entanglement evolution reflects the amplitude spatial distribution characteristics of the current input sequence;

[0094] After the initial entanglement evolution is completed, a phased probability measurement operation is performed based on the quantum state after the initial entanglement evolution. The probability distribution of the current quantum state is determined through the phased probability measurement results, and the probability value of each quantum bit after measurement is recorded at the same time.

[0095] Based on the phase difference between the phase probability measurement results and the phase difference between each quantum bit in the current quantum state, the secondary sorting sequence of the quantum bits is determined, and the secondary quantum bit recombination is performed strictly according to the secondary sorting sequence to generate a second quantum state sequence for the next round of entanglement evolution;

[0096] Repeatedly execute quantum state sequence input, real-time adaptive determination of rotation gate rotation parameters, staged probability measurement and adaptive recombination of qubits to complete multiple rounds of entanglement evolution;

[0097] After all the multiple rounds of entanglement evolution are completed, the final entangled state is used as the basis. Global quantum state cooperative measurement is performed by implementing joint projection measurement on the global basis of quantum states to obtain the joint probability distribution matrix of all qubits. Each element of the joint probability distribution matrix represents the cooperative probability value of each qubit relative to all other qubits.

[0098] The quantum state correlation strength values ​​between the initial quantum states of any two event cues are calculated based on the joint probability distribution matrix. All the calculated quantum state correlation strength values ​​are then sequentially filled into the quantum attention weight matrix according to the order of the event cues, resulting in a quantum attention weight matrix in which each matrix element corresponds one-to-one with the quantum state correlation strength between the event cues.

[0099] In this embodiment, the pre-set parameterized quantum gate is specifically:

[0100] A quantum gate sequence consisting of alternating series of rotating gates and controlled NOT gates is formed. The initial values ​​of the rotation axis direction and rotation angle parameters of the rotating gate are determined based on the centroid and standard deviation of the amplitude distribution of all qubits in the input qubit recombination sequence. The pairing order of the control qubits and the target qubits of the controlled NOT gates is determined based on the amplitude difference between adjacent qubits in the input qubit recombination sequence, sorted from largest to smallest.

[0101] In each subsequent round of entanglement evolution, the rotation axis direction and rotation angle parameters of the rotating gate are updated in real time based on the stage probability measurement values ​​obtained from the previous round of evolution. The pairing order of the control qubit and the target qubit of the controlled NOT gate is adjusted and updated in real time based on the qubit amplitude change determined by the stage probability measurement values ​​from the previous round.

[0102] In this embodiment, based on the attention weights of each event cue in the quantum attention weight matrix and the corresponding time tags of the event cue, a continuous quantum feature field model is performed on all event cuees, and a differentiable dynamic cue potential energy function is defined based on the spatial distribution of the event cuees in the quantum feature field. Specifically:

[0103] Based on the attention weight values ​​between each event cue and all other event cues in the quantum attention weight matrix, the sum of the overall attention weights of each event cue relative to all event cues is determined, and the sum of the overall attention weights is used as the initial value of the spatial density of each event cue in the continuous quantum feature field.

[0104] Using the amplitude and phase values ​​of all feature dimensions in the key semantic feature vector corresponding to each event clue as the multidimensional spatial coordinate axes of the continuous quantum feature field, the spatial position coordinates of each event clue in the continuous quantum feature field spatial coordinate system are determined based on the initial value of the spatial density corresponding to the event clue.

[0105] Based on the time tag values ​​corresponding to the event clues, the spatial location coordinates of each event clue are mapped and associated with the corresponding time tag in the spatial coordinate system of the continuous quantum feature field, thereby constructing a joint coordinate point of event clues with spatiotemporal positioning in the continuous spatial coordinate system;

[0106] Based on the spatial coordinates and time tag values ​​between the joint coordinate points of adjacent event clues, high-order interpolation operations of spatial and temporal coordinates are performed using cubic spline interpolation methods to achieve a continuous and smooth mapping of spatial and temporal coordinates between any adjacent event clues, thereby obtaining the complete spatiotemporal continuous distribution of all event clues in the continuous quantum feature field.

[0107] Based on the complete spatiotemporal continuous distribution, a local spatial density function is defined for any position in the continuous quantum feature field space. The value of the local spatial density function is determined by calculating the Gaussian weighted sum of the initial values ​​of the spatial density of all event clues within a preset radius around the corresponding position, representing the local spatial density value of the event clues at each position in the quantum feature field space.

[0108] A dynamic cue potential function is constructed using a local spatial density function. The function value of the dynamic cue potential function at any spatial coordinate position in the continuous quantum characteristic field is the negative logarithm of the local spatial density function value at the corresponding spatial position. Furthermore, the dynamic cue potential function satisfies the continuous differentiability property at any position in the quantum characteristic field space.

[0109] In this embodiment, the calculation of the local extrema of the dynamic clue potential energy function in the continuous quantum characteristic field and the corresponding potential energy gradient path, the determination of the potential clue convergence point and the event causal path, and the construction of an event correlation graph based on the potential clue convergence point and the event causal path are specifically as follows:

[0110] Within the spatial coordinate range of the continuous quantum characteristic field, based on the dynamic clue potential energy function, the multidimensional Newton iteration method is used to solve for the local extremum points in the spatial neighborhood of the joint coordinate points of all event clues, and obtain the spatial coordinate positions of the local extremum points.

[0111] Based on the derivative calculation method of dynamic cue potential energy function, the joint coordinate point of each event cue is used as the starting position. Gradient path tracing is performed in the quantum feature field space along the negative gradient direction of dynamic cue potential energy function. The local extreme point where the gradient path terminates at the spatial coordinate position is used as the potential cue convergence point of the event cue.

[0112] For the gradient path calculated from the joint coordinate points of all event clues, determine the spatial coordinate distance from the start position to the end position, retain the gradient path whose spatial coordinate distance is less than a set threshold, and determine the joint coordinate points of the event clues corresponding to the spatial coordinates of the start and end positions of the path as the start node and end node of the event causal path, respectively.

[0113] The starting and ending nodes of the event causal path are used as event nodes in the event association graph, and the directed association edges between event nodes are defined according to the gradient path direction relationship between the starting and ending nodes of the event causal path.

[0114] Event nodes and associated edges are gradually added to the event association graph until all event causal paths that meet the spatial coordinate distance threshold condition are mapped in the event association graph.

[0115] By determining the temporal order of the associated edges between event nodes through the time tag numerical relationship between the joint coordinate points of the event clues corresponding to the starting and ending nodes of the event causal path, and marking the temporal order on the associated edges, an event association graph with temporal relationship is constructed.

[0116] In this embodiment, the step of performing real-time confidence assessment of the correlation and causal relationship between events using quantum probabilistic reasoning methods based on the event association graph, historical clue paths, and newly added event clues within the current time window to obtain the event evolution path with the highest causal correlation specifically involves:

[0117] Based on the attention weight values ​​of all event clues in the historical clue paths connected to each event node in the event association graph, and the attention weight values ​​of newly added event clues in the quantum attention weight matrix within the current time window, the cumulative value of the attention weights of all connected event clues is calculated separately for each event node, and the cumulative value is determined as the initial quantum probability amplitude value of each event node, forming the initial quantum probability amplitude distribution of the event node.

[0118] The amplitude values ​​of all event nodes in the initial quantum probability amplitude distribution of each event node are mapped one by one to the qubits of the corresponding event node in the quantum probability inference space, and the amplitude value of each qubit is determined one by one, thereby generating an initial quantum probability state in the quantum probability inference space that can completely express the initial state of the event node.

[0119] Based on the directed association edges between any two event nodes in the event association graph, the attention weight value corresponding to each directed association edge is determined as the initial quantum association strength value of the association edge. Based on all the initial quantum association strength values, a quantum operator matrix is ​​constructed to express the causal relationship between event nodes. Each element of the quantum operator matrix is ​​mapped one-to-one with the attention weight value corresponding to the directed association edge between event nodes in the event association graph.

[0120] The constructed quantum operator matrix is ​​applied to the initial quantum probability state in the quantum probability inference space. The joint quantum probability amplitude distribution of each event causal path is calculated by multiplying the matrix and the quantum state to obtain the set of joint quantum probability amplitude values ​​of all event causal paths in the event correlation graph.

[0121] Perform an amplitude squaring operation on the joint quantum probability amplitude value of each event causal path to obtain a set of real-time probability values. The real-time probability value of each event causal path is the result of squaring the joint quantum probability amplitude value corresponding to the path.

[0122] The real-time probability values ​​of all event causal paths in the real-time probability value set are compared and sorted one by one. The event causal path with the largest real-time probability value is selected as the event evolution path within the current time window. The event evolution path is represented in the form of event nodes and directed edges connecting event nodes.

[0123] After determining the event evolution path within the current time window, the determined event evolution path is used as the basis for the historical clue path of the next time window. In the next time window, the initial quantum probability amplitude distribution is updated based on the newly added event clues, and the quantum probability inference step is executed again to continuously obtain the event evolution path with the largest real-time probability value in the next time window.

[0124] In this embodiment, the process of mapping and comparing the event clues corresponding to each event node in the event evolution path with the key semantic feature vectors in the original feature space, calculating the semantic feature similarity between the key semantic feature vectors, adaptively adjusting the quantum attention weight matrix based on the semantic feature similarity, iteratively performing continuous quantum feature field reconstruction and event causal path calculation until the similarity between the quantum attention weight matrix and the semantic features in the original feature space simultaneously satisfies the preset convergence condition, and outputting the final event clue evolution path, specifically:

[0125] Map each event node in the event evolution path obtained in the current time window back to the event clue corresponding to the event node in the structured data input, and obtain the original key semantic feature vector of the event clue corresponding to each event node;

[0126] Based on the key semantic feature vectors of the event nodes, the cosine similarity of the corresponding key semantic feature vectors is calculated for any two event nodes that are connected by a directed edge, so as to obtain the semantic feature similarity value between each pair of related event nodes.

[0127] The numerical difference between the semantic feature similarity value between each pair of related event nodes and the attention weight value between the corresponding event nodes in the quantum attention weight matrix is ​​calculated, and the obtained difference value is squared to obtain the set of numerical differences between the attention weight matrix and the semantic feature similarity.

[0128] The numerical difference set is summed to obtain the total numerical difference of the current quantum attention weight matrix. The total numerical difference is then compared with a preset convergence threshold to determine whether the convergence condition has been met.

[0129] When the cumulative value of numerical difference exceeds the preset convergence threshold, the attention weight value of the quantum attention weight matrix is ​​updated by scaling the attention weight value between the corresponding event nodes of the quantum attention weight matrix, using the semantic feature similarity value between each pair of related event nodes as the benchmark.

[0130] Based on the updated quantum attention weight matrix, the continuous quantum feature field modeling, dynamic cue potential energy function, potential energy gradient path calculation, event association graph reconstruction, and quantum probabilistic reasoning of event evolution path are re-executed to obtain the updated event evolution path.

[0131] Repeat the steps of semantic feature similarity calculation, quantum attention weight matrix adjustment and continuous quantum feature field reconstruction for the event evolution path until the cumulative value of the numerical difference of the quantum attention weight matrix is ​​less than or equal to the convergence threshold, and output the event clue evolution path that finally meets the convergence condition.

[0132] Example 1:

[0133] To verify the feasibility of this invention in practice, it was applied to a real-time dynamic analysis task of an intelligent emergency command system. This task requires real-time analysis of multimodal heterogeneous cues, including text, images, voice, and sensor data, to quickly identify the event evolution path and assist in decision-making. In this scenario, existing technologies generally employ classical attention mechanisms combined with traditional graph neural network methods to analyze event cues. However, due to the lack of effective nonlinear association representation and real-time dynamic iteration mechanisms, the efficiency of multimodal data fusion is low, and the accuracy of real-time analysis is insufficient to meet practical needs. Especially when event cues evolve rapidly, the analysis results show significant errors, affecting the timeliness of decision-making.

[0134] During implementation, event clue data is first generated in real time by a large-scale intelligent model, encompassing text descriptions, on-site images, voice call records, and real-time information collected by various sensors. This data is then parsed into a unified structured data format and recorded with corresponding time tags. Subsequently, feature extraction is performed on the event clue data for each modality. Specifically, event keywords and entities are extracted from text data, target appearance features and location coordinates from image data, speaker emotional features and identity information from voice data, and physical measurement parameters and state features from sensor data. After feature extraction, the system converts the obtained key semantic feature vectors into an initial quantum state representation using a quantum state encoding method. In practice, the key semantic feature vectors are normalized to the 0-1 range, and the number of qubits is determined based on the number of feature dimensions. The values ​​of each dimension of the feature vector are then mapped to the corresponding qubit amplitudes. After amplitude normalization, initial phase information is loaded, ultimately obtaining an initial quantum state representation that satisfies the physical constraints.

[0135] In the quantum state processing stage, the system uses a preset parameterized quantum gate sequence as input, with the adaptively sorted qubit recombination sequence as input. After multiple rounds of iterative entanglement evolution and quantum measurement, it obtains the quantum attention weight matrix between the clues. In the specific experiment, after the qubit amplitude and phase value sequences are adaptively arranged, they are connected in series with rotating gates and controlled NOT gates to form parameterized quantum gates. In each round of evolution, the rotating gate parameters are updated in real time based on the current amplitude centroid and standard deviation. After 5 rounds of entanglement evolution, a stable quantum attention weight matrix is ​​obtained.

[0136] Subsequently, the system models the event cues as a continuous quantum feature field based on the quantum attention weight matrix, and uses cubic spline interpolation to achieve a continuous and smooth distribution in the spatial and temporal dimensions. Based on this, a dynamic cue potential energy function is defined, and the potential energy gradient path is calculated to determine the potential cue convergence point and the event causal path, thereby constructing an event association graph that reflects the temporal relationship of events.

[0137] After constructing the event correlation graph, the system uses a quantum probabilistic reasoning method. It employs the historical clue paths corresponding to event nodes and the attention weights of newly added clues as initial values ​​for quantum probability amplitude. A quantum operator matrix is ​​then constructed to perform probabilistic reasoning, obtaining real-time probability values ​​for the event's causal path. In practical applications, after 20 rounds of dynamic time window iterations, the system consistently completes each reasoning iteration within 0.5 seconds, far lower than the average 3-second computation latency required by traditional methods, effectively meeting the high-efficiency requirements for real-time dynamic assessment of sudden events.

[0138] Furthermore, to further improve the accuracy of the analysis, the system calculates semantic feature similarity based on the original key semantic feature vectors of each event node, and adaptively adjusts the quantum attention weight matrix accordingly. It continuously iterates the reconstruction of the continuous quantum feature field and the reasoning process of the event evolution path until the cumulative difference between the attention weight matrix and the semantic similarity is lower than a set threshold of 0.001. Finally, it outputs a highly reliable event clue evolution path. In actual testing, convergence is typically achieved after 8 to 12 iterations.

[0139] To demonstrate the performance advantages of this invention, five real-world emergency scenario events were selected for evaluation results testing. For each sample, five key evaluation indicators were selected for actual measurement and prediction comparison. Detailed data results are shown in Table 1:

[0140] Table 1 Comparison of Measured and Predicted Performance Indicators for Dynamic Analysis of Event Clues

[0141]

[0142] As shown in Table 1, the proposed method excels in key metrics such as the number of event evolution nodes, average path length, judgment accuracy, attention weight error, and average computation latency. Particularly in terms of real-time computation latency, the method significantly outperforms existing classical methods in each inference iteration, with the actual measured latency consistently remaining below 0.5 seconds. Specifically, taking the S-003 sample as an example, the predicted number of nodes matches the measured value, the path length error is only 0.1 nodes, the judgment accuracy reaches 94.5%, the attention weight error is controlled at around 0.01, and the average computation latency is only 0.44 seconds. Compared to the traditional combination of graph neural networks and classical attention mechanisms, this method improves accuracy by approximately 8% and reduces latency by nearly 85%, significantly improving the real-time performance and reliability of dynamic event judgment.

[0143] Based on the above implementation results, the event clue dynamic judgment method based on large model intelligent agents proposed in this invention can efficiently and accurately process heterogeneous multimodal event clues and quickly and stably identify the dynamic evolution path of events. It shows great application value and promotion potential in emergency decision-making and risk management scenarios.

[0144] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for dynamic analysis of event clues based on large-scale intelligent agents, characterized in that, include: Acquire multimodal event cues dynamically generated by large model agents and construct structured data inputs; For each type of event clue in the structured data input, corresponding feature extraction is performed to obtain key semantic feature vectors, which are then encoded into the corresponding initial quantum state representation; The initial quantum state representation is processed using pre-set parameterized quantum gates to obtain the quantum attention weight matrix; Continuous quantum characteristic field modeling is performed for all event clues, and a differentiable dynamic clue potential energy function is defined; Calculate the local extrema of the dynamic cue potential energy function in the continuous quantum characteristic field and the corresponding potential energy gradient path, determine the potential cue convergence point and event causal path of event cues, and construct the event correlation graph; Based on the event association diagram, historical clue paths, and newly added event clues within the current time window, confidence assessment is performed to obtain the event evolution path; The event clues corresponding to each event node in the event evolution path are mapped back to the key semantic feature vectors. The quantum attention weight matrix is ​​adjusted according to the semantic feature similarity calculation results. The continuous quantum feature field reconstruction and event causal path calculation are iteratively executed to obtain the final event clue evolution path.

2. The method for dynamic analysis of event clues based on large-scale intelligent agents according to claim 1, characterized in that, The process of encoding the key semantic feature vector into the corresponding initial quantum state representation specifically involves: Perform multi-dimensional feature normalization on the key semantic feature vectors of each event clue to generate normalized key semantic feature vectors; The number of qubits required in the corresponding initial quantum state representation is determined based on the number of feature dimensions contained in the normalized key semantic feature vector, and an amplitude range is preset for each qubit, and the amplitude range is divided into multiple non-overlapping sub-intervals. Based on the pre-defined correspondence between the feature dimensions and the sub-intervals, an initial quantum bit amplitude distribution is formed; A second normalization operation is performed on the initial qubit amplitude distribution to obtain the quantum state amplitude distribution; Based on the quantum state amplitude distribution, an initial phase parameter is set for each qubit in the initial quantum state representation; According to the preset quantum bit arrangement order, the quantum state amplitude distribution and initial phase parameters are loaded one by one into the initial quantum state representation to obtain the initial quantum state representation corresponding to the key semantic feature vector.

3. The method for dynamic analysis of event clues based on large-scale intelligent agents according to claim 1, characterized in that, The initial quantum state representation is processed using pre-set parameterized quantum gates to obtain the quantum attention weight matrix, specifically as follows: Based on the amplitude and phase values ​​of each qubit in the initial quantum state representation, determine the sorting sequence of amplitude values ​​from high to low and the sorting sequence of phase values ​​from small to large for all qubits in the initial quantum state of each event clue; All qubits in the initial quantum state representation of each event clue are rearranged one by one in an alternating order of amplitude and phase to form an adaptive qubit recombination sequence; The adaptive qubit recombination sequence is input one by one into a parameterized quantum gate to perform the initial entanglement evolution; Perform a phased probability measurement operation based on the quantum state after the initial entanglement evolution, determine the probability distribution of the current quantum state through the phased probability measurement results, and record the probability value of each quantum bit after measurement at the same time. Based on the phase difference between the phase measurement results and the phase difference between each quantum bit in the current quantum state, the secondary sorting sequence of the quantum bits is determined, and secondary quantum bit recombination is performed. Repeatedly execute quantum state sequence input, real-time adaptive determination of rotation gate rotation parameters, staged probability measurement and adaptive recombination of qubits to complete multiple rounds of entanglement evolution; After completing all rounds of entanglement evolution, the joint probability distribution matrix of all qubits is obtained by performing joint projection measurements on the global basis of quantum states. The quantum state correlation strength between the initial quantum states of any two event cues is calculated based on the joint probability distribution matrix, and then filled into the quantum attention weight matrix in the order of the event cues to obtain the complete quantum attention weight matrix.

4. The method for dynamic analysis of event clues based on large-scale intelligent agents according to claim 3, characterized in that, The pre-set parameterized quantum gate is specifically as follows: A quantum gate sequence consisting of alternating series of rotating gates and controlled NOT gates, wherein the initial values ​​of the rotation axis direction and rotation angle parameters of the rotating gate are determined based on the centroid and standard deviation of the amplitude distribution of all qubits in the input qubit recombination sequence; The pairing order of the control qubit and the target qubit of the controlled NOT gate is determined by sorting the amplitude differences between adjacent qubits in the input qubit recombination sequence from largest to smallest; In each round of entanglement evolution, the rotation axis direction and rotation angle parameters of the rotating gate are updated in real time based on the stage probability measurement values ​​obtained from the previous round of evolution. The pairing order of the control qubit and the target qubit of the controlled NOT gate is adjusted and updated based on the qubit amplitude change determined by the stage probability measurement values ​​of the previous round.

5. The method for dynamic analysis of event clues based on large-scale intelligent agents according to claim 1, characterized in that, The continuous quantum characteristic field modeling of all event clues and the definition of a differentiable dynamic clue potential energy function are as follows: The sum of the overall attention weights for each event cue is determined based on the quantum attention weight matrix, and used as the initial value for the spatial density; Using the amplitude and phase values ​​of key semantic feature vectors as the spatial coordinate axes of the continuous quantum feature field, the spatial location coordinates of event clues are determined based on the initial value of spatial density. Construct joint coordinate points for event clues based on their spatial location coordinates and time tags; A cubic spline interpolation method is used to perform continuous smooth mapping of the joint coordinate points of the event clues, thereby obtaining the complete spatiotemporal continuous distribution of the event clues; Determine the local spatial density function based on the complete spatiotemporal continuous distribution; A dynamic cue potential energy function is constructed based on the local spatial density function. The function value of the dynamic cue potential energy function at any spatial location is the negative logarithm of the local spatial density function value at the corresponding location, and it satisfies the continuous differentiability property.

6. The method for dynamic analysis of event clues based on large-scale intelligent agents according to claim 1, characterized in that, The calculation of the local extrema of the dynamic clue potential energy function in the continuous quantum characteristic field and the corresponding potential energy gradient path determines the potential clue convergence point and event causal path of the event clues, and constructs an event correlation graph, specifically as follows: Based on the dynamic clue potential energy function, the spatial coordinates of the local extreme points are solved using the multidimensional Newton iteration method; Tracing the gradient path along the negative gradient direction of the dynamic cue potential energy function to determine potential cue convergence points; The starting and ending nodes of the event causal path are determined based on the starting and ending positions of the gradient path. The starting and ending nodes of the event causal path are used as event nodes in the event association graph, and directed association edges are defined between event nodes. The temporal sequence of associated edges is determined by the time tag relationship between event nodes, and an event association graph with temporal relationship is constructed.

7. The method for dynamic analysis of event clues based on large-scale intelligent agents according to claim 1, characterized in that, The process of assessing confidence based on the event association graph, historical clue paths, and newly added event clues within the current time window to obtain the event evolution path is as follows: Based on the attention weight values ​​of all event clues in the historical clue path and the attention weight values ​​of newly added event clues, the cumulative value of the attention weights of all connected event clues is calculated to form the initial quantum probability amplitude distribution. The amplitude values ​​of all event nodes in the initial quantum probability amplitude distribution of each event node are mapped one by one to the qubits of the corresponding event nodes in the quantum probability inference space, and the amplitude value of each qubit is determined to generate the initial quantum probability state. Construct a quantum operator matrix based on the directed association edges between any two event nodes in the event association graph; The constructed quantum operator matrix is ​​applied to the initial quantum probability state in the quantum probability inference space to obtain a set of joint quantum probability amplitude values. Perform an amplitude squaring operation on the joint quantum probability amplitude value of each event causal path to obtain a set of real-time probability values; The real-time probability values ​​of all event causal paths in the real-time probability value set are compared and sorted one by one, and the event causal path with the largest real-time probability value is selected as the event evolution path within the current time window.

8. The method for dynamic analysis of event clues based on large-scale intelligent agents according to claim 1, characterized in that, The process involves mapping event clues corresponding to each event node in the event evolution path back to key semantic feature vectors, adjusting the quantum attention weight matrix based on the semantic feature similarity calculation results, iteratively performing continuous quantum feature field reconstruction and event causal path calculation to obtain the final event clue evolution path, specifically as follows: Map each event node in the event evolution path obtained in the current time window back to the event clue corresponding to the event node in the structured data input, and obtain the original key semantic feature vector of the event clue corresponding to each event node; Based on the key semantic feature vectors of the event nodes, calculate the cosine similarity of the corresponding key semantic feature vectors for any two event nodes that are connected by a directed edge. The numerical difference between the cosine similarity between each pair of related event nodes and the attention weight values ​​between the corresponding event nodes in the quantum attention weight matrix is ​​calculated to obtain a set of numerical differences. The numerical differences in the set of numerical differences are summed to obtain the total numerical difference of the current quantum attention weight matrix, and the total numerical difference is compared with a preset convergence threshold. When the cumulative value of the numerical difference exceeds the preset convergence threshold, the attention weight values ​​of the quantum attention weight matrix are updated by scaling the attention weight values ​​between the corresponding event nodes in the quantum attention weight matrix. Based on the updated quantum attention weight matrix, the continuous quantum feature field modeling, dynamic cue potential function, potential gradient path calculation, event association graph reconstruction, and quantum probabilistic inference of event evolution path are re-executed to obtain the updated event evolution path.

Images