19 results about "Minimum description length" patented technology

The invention discloses a motion track construction method based on a visual large model, and relates to the technical field of computer vision, and the method comprises the steps: employing a target tracking algorithm based on Kalman filtering, and generating a smooth time sequence synchronization track from an original laser radar point cloud and an original video stream; dividing the trajectory into a series of candidate kinematic segments which are complete in kinematics and semantics by applying a hybrid method combining a minimum description length principle and visual language model semantic verification; performing preliminary semantic annotation on the segments by utilizing a visual language model to generate an initial annotation track; performing logic consistency refining on the initial labeling track until the initial labeling track is converged into a refined labeling track; and formatting the refined trajectory into a standard structured semantic trajectory representation character string. According to the method, a semantic gap between low-dimensional physical observation and high-dimensional driving intention is bridged, and ideal input is provided for understanding and prediction tasks of downstream complex scenes.

The present invention relates to a hybrid generative model device and an out-of-distribution determination method using same, wherein the hybrid generative model device comprises: a first input unit for inputting training data; a second input unit for inputting test data; a learning model unit for inputting the training data to a hybrid generative model to train the hybrid generative model, and inputting the test data to the hybrid generative model and allowing the hybrid generative model to produce an output; and an out-of-distribution (OOD) determination unit for determining OOD through a Wasserstein distance, which is measured according to the output of the hybrid generative model, between the training data and the test data and mutual information about the training data and the test data, and a minimal description length of the test data. Thereby, it is possible to effectively secure the integrity of an image as well as effectively determine in-distribution data and out-of-distribution data by using the hybrid generative model.

The invention discloses a destination reasoning method for urban vehicle tracking. The method comprises the following steps: firstly, carrying out adaptive segmentation on an original track by adopting an improved minimum description length algorithm to obtain a globally optimal track segment; then, through grid coding and a Base2vec model, the trajectory sequence is mapped into a low-dimensional distributed vector containing a geographic topological relation, namely a trajectory embedding sequence, and the sparse problem in spatial expression is relieved; then wavelet transform is applied to the track embedding sequence, and detailed information such as rapid turning or acceleration and deceleration in target movement is captured; and finally, an LSTM network is adopted, and a self-attention mechanism is integrated in the LSTM, so that the model can dynamically weight the output of different time steps according to the current trajectory segment, the network adaptively pays attention to key points in the trajectory, and the modeling capability for long-term dependence is enhanced. And finally, classifying the output of the LSTM network through a SoftMax classifier, converting a classification result into a specific geographic grid coordinate, and obtaining a predicted target destination position.

The present application belongs to the field of underwater acoustic communication signal interference elimination, and discloses a kind of underwater acoustic communication signal periodic pulse interference elimination method, device and medium, comprising: step 1: analyze the cause and characteristics of periodic pulse interference;Step 2: construct the mathematical expression of signal received by hydrophone and periodic pulse interference;Step 3: establish detection statistics according to generalized log-likelihood ratio criterion and minimum description length;Step 4: estimate the period of pulse interference;Step 5: calculate the signal of periodic pulse interference by using least square algorithm;Step 6: eliminate interference on underwater acoustic communication signal.The present application establishes the mathematical model of signal received by hydrophone and periodic pulse interference, which is much better than the traditional frequency domain filtering method.

A cell image instance segmentation method based on Bayesian inference and Boosting. For the instance segmentation of cell images, since it contains very complex image features (low foreground-background contrast, large image noise, high cell distribution density, cell adhesion, multi-layer cell overlap, etc.), it is very difficult to accurately segment each cell individual from the pathological image. Some of the methods in the past research can only segment out some cells, or require a large amount of data sets and computing power. Therefore, we urgently need a method that can completely segment each cell without consuming a lot of computing power, and the segmentation effect can achieve the expected method. The present invention introduces a multi-level set energy function and a minimum description length to perform contour reasoning, and then adds multi-layer cascade Boosting for classification and regression, so that each target in the image can be identified and accurately segmented.

The invention belongs to the field of underwater acoustic communication signal interference elimination, and discloses an underwater acoustic communication signal periodic pulse interference elimination method and device and a medium, and the method comprises the steps: 1, analyzing the generation reasons and characteristics of periodic pulse interference; 2, constructing a mathematical expression of signals received by the hydrophone and periodic pulse interference; step 3, establishing detection statistics according to a generalized log-likelihood ratio criterion and the minimum description length; step 4, estimating a pulse interference period; step 5, calculating by using a least square algorithm to obtain a periodic pulse interference signal; and 6, performing interference elimination on the underwater acoustic communication signal. According to the method for eliminating the periodic pulse interference of the underwater acoustic communication signals, a mathematical model of the signals received by the hydrophone and the periodic pulse interference is established, and the method is greatly superior to a traditional frequency domain filtering method.

This invention provides a method for early warning of borehole wall collapse risk in impact-drilled pile foundations, belonging to the field of construction risk early warning technology. This invention collects multi-source signals by deploying an ultrasonic transducer array and a mechanical vibration acceleration sensor array at multiple depth nodes within the borehole. It extracts borehole wall collapse characteristic components from strong vibration interference using empirical mode decomposition and independent component analysis. Simultaneously, it inputs the multi-dimensional time-series data of the entire borehole into a minimum description length change point detection algorithm to detect abrupt changes in borehole wall state and outputs a hazard level score. Sensor data from each depth node is input into a physically embedded spatiotemporal graph attention model to output the collapse risk probability and estimated collapse depth location. Furthermore, an adaptive learning rate adjustment function driven by a stratum change intensity index dynamically adjusts the online update learning rate of the artificial intelligence model. Finally, it integrates multi-level early warning signals to output a final early warning conclusion, solving the technical problem of the inability to provide real-time and accurate early warning of borehole wall collapse risk during impact-drilled pile construction.

The invention relates to the technical field of RAG (Retrieval Enhanced Generation), and discloses an RAG knowledge base refining system and method based on information entropy. Aiming at the problems of information overload, contradiction between efficiency and precision, module splitting and the like of an existing RAG system, four invention points of a source treatment mechanism based on a minimum description length principle, a lightweight collaborative architecture, a value-oriented retrieval strategy and a global optimization framework are provided; the knowledge value is quantitatively evaluated through information entropy, the minimum complete knowledge subset is screened, knowledge base scale compression and efficiency improvement are achieved, the method is suitable for scenes such as large language model knowledge enhancement and intelligent question and answer, and the operation efficiency and generation quality of an RAG system are remarkably improved.

The application discloses a stratum classification characteristic factor determination method, system, electronic equipment and medium, and relates to the field of oil drilling engineering.The method comprises the following steps: acquiring a historical drilling data time sequence matrix and corresponding stratum types; screening the historical drilling data time sequence matrix to obtain drilling data sets at drilling time of different stratum types; preprocessing the drilling data sets at drilling time based on a local outlier factor algorithm and a convolution smoothing algorithm to obtain preprocessed drilling data sets; determining multiple stratum classification characteristic factors in the preprocessed drilling data sets based on a minimum description length principle and Pearson correlation analysis; training a deep neural network model based on the multiple stratum classification characteristic factors of different stratum types to obtain a stratum classification model; and determining the stratum types of multiple stratum classification characteristic factors at the current drilling time by using the stratum classification model.The application improves the quality of stratum classification data.

The invention provides a medical history reconstruction system and method based on a symptom evolution graph and anti-factual reasoning, and the system comprises a symptom spatio-temporal evolution graph construction module which maps the scattered symptom description of a patient into a four-dimensional spatio-temporal coordinate system; the medical history fragment intelligent splicing engine adopts an algorithm based on a minimum description length principle to reconstruct a most probable complete medical history timeline; the active symptom inducing module is used for generating a specific action instruction or an inducible question based on the current symptom characteristics; the anti-fact diagnosis verification framework is used for evaluating the contribution degree of each symptom to the final diagnosis and identifying key symptoms and redundant information; and the symptom description standardization converter is used for automatically converting the description into standard medical terms by using a semantic mapping network based on medical ontology, and meanwhile, a nuisance mark of original semantics is reserved. Through construction of the symptom spatio-temporal evolution graph, the system can reconstruct scattered and jumping narration of patients into a complete and coherent medical history timeline, and the symptom omission rate is reduced.

The present invention discloses a risk management method and system for low-altitude aircraft based on machine learning. S1. Generate a preprocessed flight trajectory data set; S2. Apply the fractal dimension analysis method to the multi-scale flight trajectory data set to extract the complexity description vector of the flight trajectory; S3. Construct a multi-dimensional risk feature data set; S4. Automatically select the optimal risk evaluation model with the minimum description length and the best data fitting accuracy; S5. Use the selected optimal risk evaluation model to perform real-time risk assessment on the newly collected flight trajectory data set and its corresponding multi-dimensional risk feature data set to generate a flight trajectory risk evaluation output; S6. Implement dynamic risk monitoring and flight path adjustment, and provide a decision-making basis for low-altitude aircraft risk management. The present invention accurately captures high-risk feature segments with high turning rates or sharp speed changes in the trajectory, greatly improving the recognition accuracy of local risk events.

The invention provides a three-dimensional trajectory clustering method and system based on composite three-dimensional similarity distance measurement, and the method comprises the steps: obtaining three-dimensional space trajectory data, carrying out the dynamic planning segmentation based on a minimum description length principle, and extracting feature line segments representing shape features; constructing a composite three-dimensional similarity distance measurement algorithm, and calculating the geometric similarity between the feature line segments through weighted combination of a three-dimensional angle distance, a three-dimensional vertical distance, a three-dimensional parallel distance and a three-dimensional Euclidean midpoint distance; clustering the feature line segments according to the geometric similarity by using a density clustering algorithm, and outputting a clustering result; and according to the clustering result, generating a three-dimensional representative track through main direction alignment, spatial sampling aggregation and inverse coordinate transformation. According to the invention, through composite three-dimensional similarity distance measurement, the direction, position and shape features of the tracks are comprehensively captured, the tracks with similar spatial positions but different motion modes can be effectively distinguished, and the precision and robustness of three-dimensional track clustering are improved.

The present invention relates to the technical field of literature citation network analysis and graph model research, and specifically discloses a method for clustering and aggregating the core-periphery stochastic block model of a literature citation network, including the following steps: S1, defining the literature citation network; S2, constructing the core-periphery stochastic block model of the literature citation network; S3, constructing a spoke model and a hierarchical model; S4, using Bayesian inference to infer the core-periphery stochastic block model of the literature citation network; S5: Based on the principle of the minimum description length (MDL) of the model, optimizing and evaluating the core-periphery stochastic block model of the literature citation network; S6, performing structural aggregation on the core-periphery stochastic block model of the literature citation network; S7, obtaining influential literature, that is, core nodes. This solution not only can encode the prior knowledge in the literature citation network and enhance the model fitting ability by combining Bayesian inference and the stochastic block model, but also can optimize the representation accuracy of the core-periphery structure through local weighted aggregation.

The invention provides a symbol regression method and device based on a minimum description length, equipment and a medium, and relates to the field of data processing. The method comprises the following steps: acquiring real data of a plurality of first parameters (system input) and a second parameter (system output) in a target system; creating a search tree through a symbol regression method based on Monte Carlo tree search, determining MDL of a candidate formula corresponding to a node according to real data and a prediction model when the node in the search tree is expanded, updating the search tree according to the MDL, and updating the search tree according to the MDL; a single node in the search tree represents a single candidate formula (constructed by taking a first parameter as an independent variable and a second parameter as a dependent variable), and MDL represents a distance between a value of the dependent variable in the candidate formula (obtained by substituting real data of the first parameter into the independent variable of the candidate formula) and real data of the second parameter; according to the candidate formula of the minimum MDL in the created search tree, the optimal target formula expressing the relation between the first parameter and the second parameter can be obtained.

The present invention relates to the field of antenna signal processing technology, and more specifically, to a spatial-polarization domain joint search polarization generalized likelihood ratio test method applicable to arbitrary polarization-sensitive arrays and coherent and incoherent signal environments. Compared with a traditional minimum description length method, the method has higher signal source number estimation performance under low signal-to-noise ratio and low snapshot number conditions. The method models the signal source number estimation as a test problem, using 0 and 1 to represent the presence or absence of a signal. For a signal source greater than M, M signal source tests are required to complete the estimation. The method specifically comprises the following steps: calculating a likelihood ratio L1(x), then comparing the likelihood ratio L1(x) with a predetermined threshold μ1; if the value is not greater than the threshold, stopping the search, and updating the signal source number estimate #imgabs0#; otherwise, updating #imgabs1# using a maximum likelihood estimation method.

This invention discloses a network representation method and system based on a multidimensional heterogeneous resource summary view, belonging to the field of communication network technology. The method includes: acquiring network topology, device, and link status information; constructing a global original view; initializing each node and link as a supernode and superedge respectively to form an initial summary view; generating candidate supernode pairs; calculating the combined merging benefit that balances storage compression and path quality distortion; and iteratively performing supernode merging under attribute purity constraints until the summary view size meets a preset target; during the merging process, aggregating and updating the multidimensional attributes of superedges; and retaining only superedges that reduce overall storage overhead based on the minimum description length principle. This invention reduces the network state size while maintaining effective representation of multidimensional resource characteristics and possesses good dynamic adaptability, making it suitable for efficient modeling and analysis in large-scale complex network environments.

This invention proposes an FTM-corrected non-line-of-sight ranging method based on matrix-bundled CSI multipath parameter estimation. First, Channel State Information (CSI) data from WiFi devices is collected, an augmented matrix is ​​constructed, and singular value decomposition is performed to obtain a left singular matrix and singular value vectors. Second, the average power of the CSI is calculated and transformed to the logarithmic domain to determine the maximum number of candidate paths. Based on the singular value vectors, a Minimum Description Length (MDL) criterion is constructed to adaptively estimate the number of effective paths. Then, the propagation delay of each path is obtained using the matrix-bundled algorithm, a Vandermonde matrix is ​​constructed, and the received power of each path is estimated using the least squares method. Finally, the power ratio of the direct path is calculated as a reliability factor. If it is below a threshold, the second strongest path is selected from the reflection paths with power greater than the direct path. If there is only one such path, it is selected. The average delay difference between the reflection path and the direct path is calculated using power weighting, and this delay difference is used to compensate for the FTM receiving timestamp to obtain the corrected ranging distance. This invention does not require a preset number of paths, is highly adaptive, and significantly improves the accuracy of FTM ranging in non-line-of-sight environments.

The application discloses a network representation method and system based on a multi-dimensional heterogeneous resource abstract view, and belongs to the technical field of communication networks, and comprises the following steps: acquiring information of network topology, equipment and link state, constructing a global original view, and initializing each node and link as a supernode and a superedge respectively to form an initial abstract view; generating a candidate supernode pair, calculating a comprehensive income of merging which takes into account storage compression income and path quality distortion, and iteratively performing supernode merging under the attribute purity constraint condition until the size of the abstract view meets a preset target; in the merging process, multi-dimensional attributes of the superedge are aggregated and updated, and only the superedge which reduces the overall storage overhead is reserved based on the minimum description length principle. While reducing the size of the network state, the application can maintain effective representation of multi-dimensional resource characteristics, has good dynamic adaptation capability, and is suitable for efficient modeling and analysis in a large-scale complex network environment.

The application discloses a motion trajectory construction method based on a visual large model, and relates to the technical field of computer vision, which comprises the following steps: a target tracking algorithm based on Kalman filtering is used to generate smooth time sequence synchronization trajectories from original laser radar point clouds and original video streams; a hybrid method combining the minimum description length principle and visual language model semantic verification is applied to divide the trajectories into a series of candidate kinematic segments which are complete in kinematics and semantics; visual language models are used to preliminarily label the segments, and initial labeled trajectories are generated; the initial labeled trajectories are refined for logic consistency until the refined labeled trajectories are obtained; and the refined trajectories are formatted into a standard structured semantic trajectory representation string. The application bridges the semantic gap between low-dimensional physical observations and high-dimensional driving intentions, and provides an ideal input for downstream complex scene understanding and prediction tasks.