37 results about "Preprocessor" patented technology

In one embodiment, a method includes receiving data of a set of configurations of preprocessor engines, receiving measurements of performance of a device executing benchmark applications while changing a configuration of preprocessor engines selected from the set of configurations of preprocessor engines, defining an order of at least some of the configurations based on the measurements, and providing a pruned set of configurations based on the defined order of the at least some configurations.

A method and system for analyzing a vehicle winch design includes a preprocessor determining contact locations for winch components from a mesh model. The preprocessor determines load vectors at a winch assembly and counter-vectors at a fairlead based on a relative position of a winch wire and a hook retention location based on a winch load rating. A non-linear analysis system determines plastic strains, deflections, and clearances under the winch load rating based on the contact locations, load vector and counter vector, and communicates the plastic strains, deflections, and clearances to a post processing system. The post processing system compares the plastic strains to strain limit at a post processor, compares the deflection to a deflection limit, compares clearances to a clearance limit and generating a display based on comparing the plastic strains, compares the deflection to the deflection limit and compares clearances to the clearance limit.

The invention discloses an epileptic seizure detection system based on a multi-dimensional hypergraph fusion network, which comprises a multi-modal data collector, a signal preprocessor, a feature extractor, a multi-dimensional hypergraph builder, a hypergraph feature extractor, a feature fusion device, a classifier and a time sequence prediction corrector, the multi-dimensional hypergraph constructor constructs three hypergraph structures, namely an intra-modal hypergraph, an inter-modal hypergraph and a time sequence hypergraph; the feature fusion device integrates the features extracted by the three hypergraphs to generate a comprehensive feature vector; the classifier classifies the fusion features through a two-layer feedforward neural network, and outputs probability prediction of epileptic seizure; the time sequence prediction corrector applies a majority voting smoothing filter to carry out post-processing on a classification result, isolated misclassification is eliminated, and the time sequence consistency of prediction is enhanced. The invention provides an innovative multi-dimensional hypergraph fusion network, and the robustness and accuracy of epileptic seizure detection are remarkably improved by simultaneously modeling high-order relationships in modals, between modals and time sequence dimensions.

The invention relates to the technical field of ion beam physical calculation, in particular to an ion beam optical property evaluation method based on a CPU-GPU hybrid parallel stable double-gradient conjugate algorithm. According to the technical scheme, the method comprises the following steps that a CPU-GPU hybrid parallel architecture is initialized, and CPU end and GPU end memory allocation, CUDA related object creation and GPU equipment initialization and parameter setting are completed; constructing a linear equation set, a sparse coefficient matrix A and a right-end vector b, and setting an initial solution vector, convergence precision epsilon and related parameters of the number of iterations; and analyzing the sparseness of the coefficient matrix A. Through combination of three core mechanisms of CPU-GPU cooperative calculation, dynamic load balancing and intelligent preprocessor selection, efficient, stable and universal evaluation of the ion beam optical characteristics is successfully realized, and an excellent solution is provided for solving the solving problem of a large-scale sparse linear equation set in the field of high-performance calculation.

A determination system includes: a log obtainer that obtains log information; a preprocessor that generates preprocessed log information from the log information; a primary analyzer that generates primary determination result information by performing primary analysis on the preprocessed log information; a prompt generator that generates a prompt for generative artificial intelligence (AI) based on the preprocessed log information; a secondary analyzer that performs, as secondary analysis, a process of obtaining secondary determination result information by inputting the prompt to the generative AI; an overall determiner that generates overall determination result information based on the primary determination result information and the secondary determination result information; and an outputter that outputs a determination result report that includes a determination result indicated by the overall determination result information.

An apparatus for diagnosing an autism spectrum disorder (ASD) based on a graph neural network includes a preprocessor configured to acquire brain image data, designate a region of interest (ROI) of the brain, and generate preprocessed data for neural network input, a spatial feature extractor configured to extract spatial features from the preprocessed data, a temporal feature extractor configured to analyze changes in brain activity over time and extract attention-based temporal features, a spatial and temporal convergence feature unit configured to analyze spatiotemporal correlations by combining the spatial features and the temporal features, a graph generator configured to convert connectivity between regions of interest into a graph structure and implement the same as nodes and edges, and a graph classifier configured to analyze a spatiotemporal pattern of the connectivity through the graph structure and classify whether or not there is an ASD.

The invention discloses an intelligent data fusion and health early warning system and method for a multi-level health and care institution, belongs to the technical field of health and care analysis, and realizes unified modeling of multi-source heterogeneous data such as medical texts, time sequence data and images through a hierarchical heterogeneous graph neural network and a dynamic edge weight mechanism. The problem of health assessment fragmentation caused by non-standardization of data in the field of pension is solved, a space-time enhancement graph structure is introduced, a health state evolution rule is accurately captured through time attenuation factors and medical knowledge constraints, a special preprocessor is designed for special diseases, a filtering threshold value is dynamically adjusted through tremor scores, and the health assessment accuracy is improved. The problem of noise confusion is effectively solved, and the health prediction result has clinical traceability based on label system construction and graph structure visualization of the medical knowledge graph.

A Contrastive Forecasting Explanation (CFE) tool and technique provides a model-agnostic approach to forecasting explanation. The CFE tool uses an ML-based surrogate forecaster as a surrogate model. The surrogate forecaster includes a time series preprocessor, a simple concept generator, and an ML forecaster. The subsequent interpretation of the predictions of the time series forecaster is based on the behavior of the surrogate forecaster. The CFE tool interprets time series forecasts by identifying the specific temporal concepts impacting predictions and thus generates clear and reliable explanations regardless of model type. The simple concepts and predictions generated by the surrogate model are input into a perturbation-based explainer to produce feature attributions from the surrogate model. An attribution postprocessor aggregates the attributions into more coherent concepts to present a coherent, concise, and interpretable explanation.

The invention provides a data exchange visual arrangement system and method based on an API, and relates to the technical field of API data processing and flow arrangement, and the system comprises a component library which is used for storing a plurality of reusable visual components, and each visual component corresponds to a conventional operation node in a third-party platform docking process; the arrangement grouping module is used for creating arrangement groups according to API arrangement requirements associated with functions and configuring a preprocessor and a postprocessor for the arrangement groups; the visual composer is used for associating each visual component with the corresponding preprocessor and the corresponding postprocessor according to the required visual component dragged from the component library by a developer and the quoted created composing group; and according to an execution sequence and a data flow direction of each visual component defined by a developer through a connection line, completing an API arrangement process and generating a data exchange process graph. The method has the beneficial effects that the third-party docking efficiency is greatly improved, and the docking threshold is greatly reduced.

Techniques described herein relate to systems and methods for parallel DNA molecules sequencing. A preprocessor can receive raw data frames from a sensor chip including 100,000 or more cells, where each raw data frame can include detection signals from the 100,000 or more cells at a given time during the formation of the 100,000 or more cells or during the DNA molecules sequencing using the 100,000 or more cells. The preprocessor can then extract relevant information for determining states of the cells from the raw data frames, generate one or more digested frames that includes the extracted information, and send the digested frames to a processor for processing, such as base determination. Because the number of digested frames sent to the processor is less than a number of the raw data frames and the digested frames include preprocessed data, the amount of data being transferred to the processor and the amount of data processing by the processor can be reduced.

The invention discloses a calculation method for adding an irregular sub-domain without a flux boundary condition in a computational domain, which comprises the following steps of: determining an irregular sub-domain of a substrate / impurity phase added in the computational domain, and drawing the irregular sub-domain into a substrate / impurity phase shape image; processing the image into data with a required model size; establishing a calculation equation of the calculation method coupled with the irregular sub-domain added with the flux-free boundary condition in the calculation domain; performing discrete difference on the calculation equation by adopting a finite difference method to obtain a discrete difference format equation; writing a phase field preposition program by applying a discrete difference format equation; importing the data into a phase field preposition program; taking an operation result of the phase field preposition program as an initial condition of a substrate / impurity phase, and importing the operation result into a constructed phase field model for describing a dendritic crystal growth process to obtain a calculation model coupled with the substrate / impurity phase with an irregular domain; a numerical solution of the calculation model is obtained through computer simulation. According to the invention, the operation efficiency and accuracy are improved.

An input data preprocessor (IDP) and related methods for facilitating image segmentation. The preprocessor may comprise an input port (IN) for receiving an input image to be segmented by an interactive machine learning based segmentor (SEG) A subset specifier (SS) determines, based on the input image, a size specification (b) for an in-image subset. An output interface (OUT) passes the size specification to a user interface (UI) for interaction with the segmentor (SEG). The proposed input data preprocessor (IDP) may preferably be used in interactive segmentation, to reduce the number of iteration cycles.

This application relates to a method and system for solving conjugate gradient finite element models based on sparse convolution preprocessing. The method includes establishing a structural finite element model, generating a structural stiffness matrix A and a load vector b, and constructing a linear equation system Ax=b, where A is a sparse symmetric positive definite matrix. A preprocessing sub-generator is constructed by training different structures using a sparse convolutional neural network. The stiffness matrix A is input into the preprocessing sub-generator to obtain a preprocessing factor. A symmetric positive definite preprocessor is constructed based on the preprocessing factor. The linear equation system Ax=b is solved using the preprocessed conjugate gradient method to obtain the displacement response vector x. This application optimizes the condition number of the preprocessed matrix to improve convergence speed and reduce solution time. It adapts the sparse convolutional U-net structure to large-scale sparse matrices, improving training efficiency and forming a unified and scalable preprocessing framework for structural engineering, providing a general and efficient preprocessing strategy for large-scale finite element model analysis.

Techniques described herein relate to systems and methods for parallel DNA molecules sequencing. A preprocessor can receive raw data frames from a sensor chip including 100,000 or more cells, where each raw data frame can include detection signals from the 100,000 or more cells at a given time during the formation of the 100,000 or more cells or during the DNA molecules sequencing using the 100,000 or more cells. The preprocessor can then extract relevant information for determining states of the cells from the raw data frames, generate one or more digested frames that includes the extracted information, and send the digested frames to a processor for processing, such as base determination. Because the number of digested frames sent to the processor is less than a number of the raw data frames and the digested frames include preprocessed data, the amount of data being transferred to the processor and the amount of data processing by the processor can be reduced.

To provide a video encoding device, a pre-processing device, and a video encoding method that improves encoding efficiency.SOLUTION: A video encoding device 100 includes a pre-processing unit 110 that performs pre-processing by a first neural network on video signals of a plurality of frames including an input video frame by using a first picture reference structure, and outputs a pre-processed video signal. The video encoding device 100 also includes a video encoding unit 120 that encodes the pre-processed video signal using the first picture reference structure.SELECTED DRAWING: Figure 1

The invention provides a multi-component periodic linear frequency modulation signal processing method and system based on deep learning, and the method comprises the steps: introducing a double-flow multi-task deep learning network as an intelligent preprocessor, and converting a conventional blind estimation problem into a prior-guided precise estimation problem; the method fundamentally overcomes the high dependence of an existing method on key prior information such as the number and period of components, and remarkably improves the decoupling performance and parameter estimation robustness of weak PLFM signals in the environment of extremely low signal-to-noise ratio, strong pulse noise and complex multi-component interleaving. Meanwhile, through efficient combination of a time-frequency mask denoising function of the deep learning network and subsequent SVD-HPD-DPT processing, high-precision separation and parameter extraction of multi-component signals are realized on the premise of keeping relatively low calculation complexity, and the contradiction among real-time performance, anti-interference performance and estimation precision of a traditional method is effectively solved.

The application discloses a dynamic link library reflection method based on a C language preprocessor, and comprises the following steps: S1, acquiring the file name of a dynamic link library to be agented and a function list to be agented, wherein the function list comprises the function name, return type and parameter type of each function; S2, constructing a registerer by using the macro definition of the C language, wherein the input of the registerer is the file name of the dynamic link library, the function name to be agented, the return type and the parameter type; S3, generating an agent function by using the registerer in the preprocessing process of the compiler at the compiling stage; S4, registering the function to be agented in the source code by using the registerer in step S3; and S5, calling the function to be agented in S1 in the business, and compiling and running the software together with the registration code in step S4. The application solves the compatibility problem caused by using different version library modules or using the same library module in different versions of the same product.

A weak signal detection method based on stochastic resonance includes: a raw signal acquisition unit acquiring raw signals and sending the acquired signals to a preprocessor; the preprocessor using preprocessing methods for different weak wireless signals in turn, and then sending the processed information to a stochastic resonance adaptor; the stochastic resonance adaptor using stochastic resonance adaptive enhancement methods for different weak wireless signals in turn, and determining the signal type and channel based on the results; and a high-power Wi-Fi signal generator controlling the Wi-Fi signal to yield the spectrum corresponding to the weak wireless signal being transmitted, thereby enabling the co-transmission of various weak signals and Wi-Fi signals.

This invention discloses a graph neural network accelerator and device based on topology reconstruction and dynamic early retirement. The graph neural network accelerator includes an interconnected graph preprocessing module and an accelerator body. The graph preprocessing module includes a feature preprocessor, a pruner, a community divider, a topology optimizer, and an encoder to generate optimized graph structure data in ECSR format. The accelerator body includes an aggregation engine, a combination engine, a hierarchical storage unit, and a global configuration controller. The combination engine includes a confidence evaluation unit for evaluating node convergence confidence during graph neural network inference to support dynamic early retirement acceleration. This invention aims to address the low inference efficiency of general-purpose CPUs and GPUs due to the sparse and unstructured hybrid computing characteristics, as well as the shortcomings of existing graph neural network accelerators in topology adaptability, computational resource utilization, and model support, thereby improving the inference efficiency of graph neural network accelerators.

The application relates to the technical field of radar target detection, and discloses a forest scatterer type identification and positioning system based on radar polarization decomposition, which comprises a polarization feature preprocessor, a main controller and a threshold calibration unit; the main controller performs asymmetric scheduling and review; the threshold calibration unit uses the clutter echo obtained through the review to calculate the background clutter feature baseline in real time, and dynamically updates the classification threshold used by the polarization feature preprocessor; the application uses the review information flow to realize closed-loop self-calibration of the classification threshold, makes the system free from the dependence on the priori of the static environment, can automatically track the feature drift of the background clutter caused by meteorological changes, and avoids the risk that the scheduling logic is disabled due to the saturation of the classifier.

A computer-implemented method of training a neural network to recognize sensory patterns includes obtaining input data, preprocessing the input data in one or more preprocessors of the neural network, and applying the preprocessed input data to a core portion of the neural network. The core portion of the neural network includes a plurality of principal neurons and a plurality of interneurons, and is configured to implement a feedback loop from the interneurons to the principal neurons that supports persistent unsupervised differentiation of multiple learned sensory patterns over time. The method further includes obtaining an output from the core portion, and performing at least one automated action based at least in part on the output obtained from the core portion. The neural network may be adaptively expanded to facilitate the persistent unsupervised differentiation of multiple learned sensory patterns over time by incorporating additional interneurons into at least the core portion.

A platform for node-based core operations and trust data asset management incorporates data integrity preservation via trust controls and includes virtual machine(s) configured and deployed to execute an operating system to access the computing platform via a virtual network. The computing platform includes a blockchain and interconnected data processing nodes each including functional layers with an asset management service layer to establish a blockchain computing function that preserves data integrity by defining a single view of trust data assets referenced in transaction(s) from authoritative source(s). The preserving applies identification, standardization, and normalization rules in validation smart contracts of an ingestion preprocessor to establish the single view and executes registration and certification processes that incorporate smart contract trust controls to enhance data integrity. An API framework is maintained to support master and reference data management data processing to establish node-based core operations and trust data asset management under trust control.

Systems and techniques are described herein for training a video preprocessor. For instance, a method for training a video preprocessor is provided. The method may include processing training video data using a video encoder-decoder to generate intermediate video data; processing the intermediate video data using the video preprocessor to generate output video data; determining a loss based on the output video data and the training video data; and adjusting parameters of the video preprocessor based on the loss, wherein the video preprocessor is configured to process video data to generate preprocessed video data and to provide the preprocessed video data to a video encoder.

An object of a system according to an embodiment is to realize self-optimization and automatic adjustment in an operation of a network.SOLUTION: In one embodiment, a system comprises a AI collector, a pre-processor, a generative data trainer, a monitor, a feedback loop, a self-optimizer, and a model improver. The data collection unit collects data. The preprocessing part preprocesses the data collected by the data collection part. The generative AI trainer is configured to train the generative AI using the output data preprocessed by the preprocessor. The monitoring unit monitors the network using the generated AI trained by the generated AI training unit. The feedback loop unit performs feedback based on the operation data of the network monitored by the monitoring unit. The self-optimization unit performs self-optimization based on the feedback obtained by the feedback loop unit. The model improvement unit improves the model based on the data obtained by the self-optimization unit.SELECTED DRAWING: Figure 1

This disclosure provides an interactive control method, device, and electronic device for a shuttle frame. The shuttle frame interface displays a target list and a source list. When a data shuttle operation is received for a data item in the source list, the selected data item is moved from the source array to the target array, and a calculation effect is applied to the data item based on a preset calculation logic. When a data shuttle operation is received for a data item in the target list, the selected data item is moved from the target array to the source array, and the applied calculation effect is removed. The shuttle frame interface is updated based on the source and target arrays after the data migration. This provides an intelligent shuttle frame that supports dynamic calculation, enabling simultaneous data allocation and calculation. It transforms the shuttle frame component from a single-function data selector into a data preprocessor, making it suitable for complex business scenarios requiring data pre-calculation.

This application discloses a method and apparatus for processing multimodal data. Relating to the field of financial technology, the method includes: acquiring initial multimodal data carrying metadata tags, the initial multimodal data including text data, image data, and audio data; allocating the initial multimodal data to different types of preprocessors based on the metadata tags to obtain processed multimodal data; mapping the processed multimodal data to various nodes in a target knowledge graph according to preset constraints; adjusting the multimodal feature weights of each node in the target knowledge graph relative to its neighboring nodes based on the stage-specific business characteristics of each business stage to obtain differentiated multimodal features adapted to the current business scenario and each business stage; and obtaining the business processing result by inputting the differentiated multimodal features into a target decision model. This application solves the technical problem of low processing efficiency for multimodal data in related technologies.

This invention belongs to the field of electromagnetic simulation calculation and provides an electrically large-scale electromagnetic simulation method based on a hybrid algorithm for partitioned solution. First, the large-scale electromagnetic target is divided into solution regions according to component structure. For different algorithm solution domains, a domain decomposition method is used to decompose the hybrid algorithm solution matrix into subdomains, obtaining the hybrid algorithm solution matrix equation after domain decomposition. Then, for the hybrid algorithm solution matrix after domain decomposition, an outermost preprocessor is constructed using a block Jacobi preprocessing method, and an inner preprocessor is constructed using a multi-wavefront block incomplete decomposition method. Finally, multi-threaded parallel technology is used to iteratively solve the hybrid algorithm solution matrix equation, obtaining solutions for both the finite element solution domain and the integral equation solution domain. Based on this, this invention has advantages such as significant multi-algorithm fusion, high preprocessing efficiency, small memory footprint, and good parallel scalability, enabling efficient electromagnetic simulation analysis of electrically large-scale targets.

The invention discloses a high-precision 3D defect measurement method and device, and the method comprises the following steps: generating an initial depth map and a confidence map through phase demodulation and depth reconstruction based on an original optical signal collected by a 3D camera; discretizing the Poisson equation by using a finite difference method, constructing a weighted gradient field and generating a sparse linear equation; graph structure features are extracted through a graph convolutional neural network, a dynamic preprocessor matrix is predicted, and the solution stability is improved; a generalized minimum residual algorithm is combined, matrix partitioning, mixed precision and parallel computing acceleration solution are adopted on a graphics processor, and a high-precision optimized depth map is obtained; and positioning an abnormal region through gradient analysis, threshold segmentation and morphological filtering, extracting a contour and mapping the contour to a three-dimensional space, and generating a complete defect detection result containing position, size and visual labeling. According to the invention, the accuracy of 3D camera defect detection is improved.

In an example embodiment, semantic chunking is combined with a metadata enrichment mechanism where chunks are enriched with additional metadata prior to being embedded. These embeddings may then be stored in a vector database and used to perform enriched similarity searches for augmentation of context provided to an LLM during LLM generation requests.