17results about How to "Efficient modeling" patented technology

The invention relates to the technical field of water chilling units, in particular to a central air conditioner water chilling unit energy efficiency ratio prediction method based on XGBoost and LSTM, a medium and equipment, and the method comprises the following steps: S1, an initial training stage dominated by simulation data; s2, a dynamic weight adjustment stage: starting to accumulate real data along with the operation of the water chilling unit; and S3, an incremental learning and model updating stage: continuously monitoring a prediction error of the model on new real data, and when the error exceeds a threshold value or enough new data is accumulated, triggering an incremental learning mechanism. According to the method, modeling capabilities of simulation and real data are fused, and coverage and physical constraint characteristics of simulation data are effectively utilized; structure-time sequence feature fusion is realized by combining the static feature modeling capability of the XGBoost and the time sequence mode capturing capability of the LSTM; the method is suitable for long-term deployment and iterative optimization in an industrial environment.

This invention discloses a drug target affinity prediction method that integrates PPI quality and uncertainty. The method constructs a drug molecule map and a multimodal protein structure representation, and extracts multi-source features by combining the local PPI sub-map of the target protein. By calculating the protein's low-frequency level, prediction uncertainty, and PPI quality, a PPI quality-aware gating factor is generated to adaptively adjust the PPI information injection intensity, and a residual enhancement strategy is used to preserve the original protein features. Subsequently, the drug representation and the enhanced protein representation are fused using adaptive gating, and the result is input into a prediction network to output the drug-target affinity. This method effectively integrates protein function and interaction information, improves the prediction stability of low-frequency proteins and the model's generalization ability, and provides an accurate and reliable computational tool for drug screening and candidate molecule selection.

This invention discloses a video summarization method based on two-layer routed sparse attention and spatial pixel recalibration, belonging to the field of computer vision. The method includes: reading the input video and extracting frame-level feature vectors; constructing a video summarization generation model, performing channel enhancement on the frame-level feature vectors to obtain enhanced features; performing two-layer routed sparse attention and spatial pixel recalibration on the enhanced features, and fusing them to obtain fused features; inputting the fused features into an importance scoring regression network, outputting frame importance scores, and selecting frames to generate video summaries. This invention effectively fuses global dependencies and local details by combining two-layer routed sparse attention and spatial pixel recalibration, accurately identifying key segments in the video while reducing computational complexity. Experimental results based on the SumMe and TVSum benchmark datasets demonstrate that the method of this invention exhibits good performance in video summarization tasks.

The invention relates to the technical field of cranial neuroscience, and provides a model construction method, an image processing method, equipment, a medium and a product. The method comprises the following steps: acquiring representation data corresponding to a brain medical image and a classification label of the brain medical image for an Alzheimer's disease scene; sequentially constructing a plurality of processing layers under the scene of the Alzheimer's disease based on the representation data and the classification labels; constructing a Laplacian matrix corresponding to the initial classification model, constructing an objective function of the initial classification model based on the initial classification model and the Laplacian matrix, and solving through the objective function to obtain an output weight of the initial classification model; and updating the initial classification model based on the output weight to obtain a target classification model for the Alzheimer's disease scene. Through the technical scheme of the invention, efficient modeling in a small sample scene is realized, the problems that a deep learning model depends on a large number of samples and iteration time is long are avoided, and the classification efficiency is improved.

The application discloses a kind of reconfigurable intelligent surface auxiliary wireless environment modeling method based on three-dimensional Gaussian splash technology, belong to wireless communication field, this method will three-dimensional Gaussian splash technology from computer graphics field innovatively migrate to RIS auxiliary wireless communication system modeling field, the explicit parameterization thought of 3D-GS is applied to radio frequency electromagnetic field modeling, break through the limitation of traditional deep learning "black box" modeling;Adopt two-stage joint modeling framework: for the characteristics of RIS system, "TX→RIS" and "RIS→RX" cascade modeling strategy is designed, can accurately capture the electromagnetic regulation effect of RIS;With explicit physical parameterization: replace large-scale neural network weight with Gaussian parameter with clear physical meaning, realize light weight;In addition, by constructing complete differentiable rendering process, support efficient gradient descent optimization, ensure training efficiency and convergence.

The application relates to a product popular trend prediction method based on zero sample learning and belongs to the technical field of data analysis and prediction. The method comprises the following steps: acquiring image data and text data of a product and respectively converting the data into a visual feature vector and a semantic vector; generating a style label set according to the two vectors and constructing style category semantic features; acquiring time-series heat data of each style category of the product and constructing a heat evolution sequence of each style category of the product; extracting trend features of the sequence and optimizing the trend features by adopting a time guide mechanism to obtain optimized time-series trend features; constructing a weighted adjacency matrix according to clustering results of each style category of the product; performing multi-order information diffusion processing on the matrix to generate graph features; and fusing the style category semantic features, the optimized time-series trend features and the graph features to obtain comprehensive features of each style category of the product, so that a popular trend result of the product is obtained. The application improves zero sample recognition capability and product popular trend prediction accuracy.

The invention belongs to the technical field of biology, and relates to a construction method and application of a mouse exogenous gene expression model based on an adenovirus vector. The exogenous gene is delivered through the adenovirus vector, the exogenous gene is subjected to site-specific integration to the specific site of the mouse target cell by using the double recombinase, the efficiency is high, the integration site and copy number are controllable, the safety is good, various function acquired variations can be efficiently modeled, and the method can be applied to establishment of a tumor model and research of precision medical treatment.

A segmented modeling and simulation method for quantum well semiconductor optical amplifiers based on spatiotemporally staggered half-step partitioning is proposed. The method includes extracting and preprocessing parameter data of the quantum well semiconductor optical amplifier device and the quantum well electro-optic gain material; and performing stepwise segmentation of the optical amplification direction of the active region of the quantum well semiconductor optical amplifier using spatiotemporally staggered half-step partitioning. This method constructs a high-speed and efficient simulation design through novel spatiotemporally staggered half-step partitioning modeling and calculation, which can improve the efficiency of large-signal time-domain simulation calculation of devices such as quantum well semiconductor optical amplifiers and quantum well semiconductor lasers and reduce the time consumption of optimization design.

The application discloses a weather forecasting method and device, a computer program product and an electronic device. The method comprises: obtaining historical meteorological features for meteorological prediction of a to-be-predicted area; using a plurality of hierarchical transformer modules to analyze meteorological physical laws of the historical meteorological features layer by layer to obtain fused meteorological features of each layer, wherein each transformer module is used to guide an attention mechanism to focus on local meteorological features of the to-be-predicted area and its neighborhood according to a gating position encoding to obtain first global meteorological features, and correct each local meteorological feature in the first global meteorological features according to a global spatial position bias term to obtain second global meteorological features; and using a multi-scale output head module to fuse the fused meteorological features of the plurality of layers to obtain a predicted upper air field and a predicted ground field of a prediction period. The application solves the technical problem of large prediction error of an existing weather forecasting model.

This invention discloses a method for 3D environmental feature extraction and electromagnetic propagation path modeling based on neural networks, belonging to the field of channel modeling technology. The method includes: S1: acquiring wireless channel data of the target scene and corresponding 3D spatial point cloud environmental data; S2: obtaining a physical environment feature dataset for neural network training; S3: constructing a 3D environmental feature extraction model to generate implicit environmental feature representations reflecting the relationship between environmental spatial distribution and material properties; S4: constructing an electromagnetic propagation direction prediction model and a power prediction model to predict the step-by-step evolution of the electromagnetic propagation path and output electromagnetic propagation path-level characteristics. This invention solves the problems of complex and inefficient existing modeling processes, difficulty in directly utilizing measured point cloud data, and difficulty in efficiently generating multipath propagation paths. This invention improves modeling efficiency and reduces labor costs, enabling direct modeling of point cloud geometry and material properties while reducing the computational complexity of multipath path inference.

The invention discloses a three-dimensional geologic structure modeling method and system based on SH-LDM and a medium, and relates to the technical field of deep learning and geologic modeling crossing, and the three-dimensional geologic structure modeling method based on SH-LDM mainly comprises the steps: carrying out the preprocessing of original data, and obtaining a training data set and a test data set; training the multi-branch self-encoding network by using the first-stage loss function, and training the multi-condition diffusion network by using the second-stage loss function; and according to soft data and hard data of a real scene, utilizing the trained multi-branch self-encoding network and the multi-condition diffusion network to obtain a three-dimensional geologic structure generation image. By implementing the SH-LDM-based three-dimensional geologic structure modeling method and system and the medium provided by the invention, high-precision and efficient modeling of a complex non-stationary underground structure can be realized.

The invention discloses an indoor line series fault arc detection method based on multi-feature fusion, and the method comprises the steps: collecting a current signal at a main line of an indoor line, carrying out the normalization preprocessing of the current signal, and extracting a time domain, a frequency domain and multiple types of features of the time domain and the frequency domain, so as to comprehensively reflect the non-stationary features of the current signal. For samples under different working conditions, redundant features are removed by using an improved feature selection algorithm in combination with Pearson correlation analysis, and a final feature set with strong discrimination capability is constructed. Furthermore, an extreme learning machine model is introduced, rapid modeling and efficient classification are achieved based on a random initialization weight and a generalized inverse solution method, and high detection precision and real-time performance can be kept under complex working conditions and noise interference. The method has the advantages of being comprehensive in feature expression, efficient in model training, high in detection accuracy and high in robustness, the missing report rate and the false report rate of the indoor line series fault arc can be effectively reduced, and the safety and the reliability of an electrical system are improved.

The embodiment of the application relates to the technical field of software engineering and artificial intelligence, and discloses a process modeling method for human-machine cooperation and related equipment, the method comprising: constructing a hierarchical recursive process model, the process model comprising a process domain layer, an activity layer and a task layer; using an eight-tuple model to formally describe each task, the eight-tuple model comprising a task identifier, an input artifact set, an execution logic, an output artifact set, a role performer, a constraint set, a cognitive level and a deterministic classification, defining four types of role performers for the process model, including a human role, an agent role, a system role and a hybrid role, and based on the four types of role performers and the eight-tuple model, rendering each activity of the process model as an activity graph on a preset workbench interface through differentiated visual lanes, and in this way, the embodiment of the application effectively realizes modeling of AI tasks under the framework of organizational system constraints and standard specifications in a human-machine cooperation scenario.

This invention discloses a lemon health analysis method, relating to the fields of intelligent agricultural monitoring and plant health diagnosis. It involves acquiring continuous images of lemon plants using a hyperspectral camera, and then using these images to detect changes in a dynamic differential spectral segmentation model. Sequence modeling is performed on the pixel-level multi-band spectral data. An adaptive motion-aware scheduler automatically suppresses loss when the proportion of motion change mask pixels exceeds a preset dizziness threshold to avoid background misjudgment, generating a pixel-level dynamic change map. This is compressed into a pathological dynamic state vector using a variational autoencoder, and end-to-end training is performed with action-conditional video prediction as the training objective. In the extrapolation phase, a recurrent dynamic model is used to perform multi-step imaginative prediction based on candidate action sequences, outputting the future evolution trajectory of the pathological dynamic state vector and generating a quantitative early warning or decision simulation report on the outbreak risk level. The system also includes change type classification and model closed-loop optimization functions.

The invention discloses a coal mine rock burst risk space-time early warning method and system based on multi-source image data fusion, a storage medium and equipment. The method comprises the following steps: 1, collecting multi-source heterogeneous data; 2, data preprocessing; 3, constructing a gridding cloud picture of each type of data; 4, constructing a rock burst early warning model of an encoder-decoder network structure based on the MSA-Unet; 5, performing model training and optimization; 6, dynamically updating the early warning area and the data; 7, performing end-to-end reasoning to generate a rock burst risk early warning cloud picture; and 8, establishing a circulating early warning mechanism taking time or footage distance as a triggering condition. Multi-source heterogeneous data is converted into unified image representation, the problem of fusion and utilization of multi-source data such as rock burst disaster geological data and micro-seismic monitoring data is solved, and early warning and assessment of rock burst danger in an advanced area of a mining working face are realized based on time-space division of the data. And space-time dynamic early warning of the rock burst risk of the coal mining working face is realized.

The application discloses an encryption flow attack detection method and system based on a pulse nerve and a bidirectional LSTM, and comprises the following steps: obtaining encryption flow data, performing flow-level reorganization and pulse coding, and generating a pulse input sequence; inputting a multi-layer pulse nerve network, calculating a local loss by using a DECOLLE local learning mechanism, and completing network training to generate a hierarchical pulse event representation sequence; calculating a change amplitude and detecting an event trigger boundary, and constructing an intra-segment and segment-level representation sequence; constructing a double-channel input sequence, inputting a bidirectional LSTM network to generate a global context gating vector; performing gating weighting and inter-layer consistency constraint based on the context gating vector, constructing a joint training target, and synchronously updating network parameters; obtaining to-be-detected encryption flow data, and generating an encryption flow attack detection result based on the trained model. The application improves the stability and accuracy of encryption flow attack detection.

The invention relates to a product fashion trend prediction method based on zero sample learning, and belongs to the technical field of data analysis and prediction. Comprising the following steps: acquiring image data and text data of a product, and respectively converting the image data and the text data into visual feature vectors and semantic vectors; generating a style label set according to the two vectors, and constructing style category semantic features; obtaining time sequence popularity data of each style category of the product, and constructing a popularity evolution sequence of each style category of the product; extracting trend characteristics of the sequence, and optimizing the trend characteristics by adopting a time guide mechanism to obtain optimized time sequence trend characteristics; according to the clustering result of each style category of the product, constructing a weighted adjacent matrix; performing multi-order information diffusion processing on the matrix to generate map features; and fusing the style category semantic features, the optimized time sequence trend features and the map features to obtain comprehensive features of each style category of the product, thereby obtaining a popularity trend result of the product. According to the invention, the zero sample identification capability and the product popularity trend prediction accuracy are improved.