218 results about "Data quantization" patented technology

The invention relates to the technical field of image recognition, in particular to an image monitoring system for traditional village heritage risk assessment, and the system comprises a heritage image collection module which is used for obtaining an image data stream of a target traditional village building surface, carrying out the geometric correction of original heritage image data, carrying out the image brightness equalization processing, and obtaining an image data stream of the target traditional village building surface; and establishing a calibrated image set. According to the method, image distortion and local overexposure caused by shooting angle difference or uneven illumination on the surface of a traditional village building are eliminated through geometric correction and brightness equalization processing, and the input data quality of subsequent feature analysis is improved. The consistency degree of crack textures in a neighborhood range is quantified based on gray gradient direction field data, the gray contrast of continuous crack edges is enhanced in combination with a dynamic threshold adjustment mechanism, non-structural texture interference is inhibited, and separability of micro cracks and background materials is enhanced.

The invention discloses a self-correction structured bid evaluation guide method for public resource transaction, and relates to the field of resource transaction processing, the method is based on a bid evaluation system, and the method comprises a data preprocessing unit, a structured analysis unit, a rule unit, an intelligent evaluation unit, a workflow management unit, a report generation unit and a data filing unit. Historical similar project scoring data is analyzed based on a large language model through an intelligent review unit, a real-time scoring reference interval is generated, expert experience and data driving are combined, professional judgment is reserved, scoring randomness is reduced through data quantification, a workflow management unit adopts an LSTM neural network to construct a review behavior prediction model, and the review behavior prediction model is used for predicting the review behavior. According to the method, the degree of deviation between expert scores and predicted values is compared in real time, meanwhile, multi-party review cases are integrated based on federal learning to form a dynamic knowledge base, review after the event is changed into real-time monitoring and intelligent intervention, the objectivity of bid evaluation is enhanced, and subjective item scores tend to be standardized.

Workloads for executing a graph neural network (GNN) may be divided among various processing units, such as a central processing unit (CPU) and a neural processing unit (NPU). The NPU may include a data processing unit (DPU) and a digital signal processor (DSP). The CPU may perform precomputation, model optimization, hardware optimization, and compilation. For example, the CPU may precompute a parameter matrix and use the parameter matrix as internal parameters of a GNN. The CPU may also perform node padding, approximation computation, or transfer of DSP operations to DPU to optimize the GNN. The CPU may also perform sparsity data compute and storage, vertical fusion of DSP operations and DPU operations, or data quantization to optimize performance of the NPU. The compiled GNN may be provided to the NPU, and the DPU and DSP may perform the operations in the compiled GNN to produce a prediction of the GNN.

The invention relates to the technical field of photovoltaic power station management, and discloses a photovoltaic power station Internet of Things intelligent management and cloud edge collaborative optimization system. The system comprises an operation data acquisition module which is used for acquiring power station data in multiple dimensions, quantitatively identifying an abnormal operation state and generating a management early warning signal; the equipment state evaluation module is used for extracting equipment characteristics and dynamically evaluating the self-maintenance capability to obtain a state evaluation reference value; the edge processing module is used for monitoring parameters such as data traffic and predicting and analyzing edge processing efficiency to obtain an edge calculation evaluation value; the cloud edge collaboration module is used for jointly analyzing collaborative optimization efficiency and generating a regulation and control signal; the optimization parameter generation module is used for generating equipment maintenance and computing power distribution parameters; the system further comprises an authority control module and a dynamic adjustment module. According to the system, intelligent management and cloud edge collaborative optimization of the photovoltaic power station are realized, the operation efficiency and reliability of the power station are improved, and the safety and dynamic adjustment capability are enhanced.

The invention discloses an image recognition and correction system based on an optical model and environment perception, and relates to the technical field of image processing, and the system comprises an environment perception module which is used for collecting multi-medium refractive indexes, interface three-dimensional forms and initial image data through a refractive index sensor, a three-dimensional laser scanner and a visual camera; an environment perception data set is formed through dynamic correction and multi-source fusion; the optical model is constructed on the basis of environment sensing data, the light propagation path is calculated in combination with the interference compensation sub-model, and the interference compensation sub-model can quantify the interference influence by analyzing related data acquired by the environment sensing module for interference factors such as impurities and temperature gradient in a medium, so that the interference compensation precision is improved. The compensation term is generated and fused into the light propagation equation, so that the light propagation path is corrected, the calculation accuracy of the light propagation path is ensured, a firm and reliable data support is provided for subsequent image recognition and correction, and the precision and quality of image recognition and correction are improved.

The invention relates to the technical field of operation management, in particular to a resource optimization-oriented operation activity data quantitative analysis method and system, and the method comprises the steps: obtaining multi-modal data to construct enterprise operation digital twins; on the basis of the twin data, inferring a causal relationship between variables through a causal discovery algorithm, quantifying a causal effect by using a graph neural network, and constructing a high-fidelity simulation environment on the basis; and training and generating an optimal resource allocation strategy model in the environment by adopting a deep reinforcement learning engine. In the decision-making stage, causal intervention simulation is carried out on the candidate schemes through a do operator, so that the net effect of the candidate schemes is evaluated, and the optimal scheme is selected. And finally, the scheme is deployed and continuously monitored, and a new data iteration optimization model is utilized to form a self-adaptive closed loop of'data-insight-decision-optimization '. According to the resource optimization-oriented data quantitative analysis method and system, the problems that a decision-making link is split and depends on artificial experience in the prior art are solved.

The invention discloses a PUF (Physical Unclonable Function) structure based on delay difference quantization, and the structure comprises a PUF circuit which comprises two time delay circuits which are connected in parallel; a plurality of layers of differential tapped circuits TDC, and decoders connected with each layer of differential tapped circuit TDC; the delay data quantization unit is connected with the multi-layer decoder, and the delay data quantization unit is connected with the LFSR confusion structure; the two delay circuits are connected with the multi-layer differential tap circuit TDC through the signal selection structure, the signal selection structure inputs delay signals which respond first and then respond in the two delay circuits into a start end and a stop end of each layer of differential tap circuit TDC, the decoder determines delay time of the two delay circuits under corresponding delay resolution, and the decoder outputs the delay time to the multi-layer differential tap circuit TDC. The delay data quantization unit is used for averaging the delay time under different time delay resolutions to serve as final delay data, and the final response R is output after the final delay data is confused by the LFSR confusion structure. Signals of the two symmetrical delay paths are respectively input into a framework of a multi-layer TDC quantization circuit, so that delay difference digital code quantization is realized.

The invention provides a parallel world simulation resource dynamic scheduling optimization method and system. The method comprises the following steps: detecting a dependency relationship between tasks in parallel world simulation resources to obtain topological data, and quantifying resource competition intensity and sequence constraint to form an associated feature matrix; a cross-dimension optical signal transmission path is constructed through dynamic change of a programmable photon array analysis matrix, and a low-conflict resource transmission channel map is formed; synchronously processing the matrix and the channel map to generate a global task execution sequence; executing a multi-world resource occupancy operation and generating a resource occupancy relation graph; and calculating a cross-dimension resource conflict probability in the atlas, and dynamically adjusting a task execution sequence to avoid a simulation deadlock risk. According to the method, nanosecond-level dynamic resolution of cross-dimension resource conflicts is realized, the simulation deadlock rate is zeroed, and the task throughput is improved.

The invention provides KV cache data quantization equipment, which comprises a precision decision interface, a compressor and a storage interface, and is characterized in that the precision decision interface is used for determining a quantization precision mark of each token in combination with token quantization difficulty, and generating a mixed quantization precision instruction according to the quantization precision marks corresponding to all the tokens; the compressor is used for acquiring a floating point Key value and / or a floating point Value value corresponding to each token, quantizing the floating point Key value and / or the floating point Value value based on a mixed quantization precision instruction issued by the precision decision interface, and generating metadata with a scaling factor and a zero point of each token and corresponding quantization data; and the storage interface is used for combining the quantized data corresponding to the quantized precision marks with the low-precision bit width in pairs, associating the quantized data with the metadata through an address mapping table, and separately storing the quantized data and the metadata. According to the invention, the storage space occupied by the KV cache can be reduced, and the reasoning speed of the model is improved.

The invention provides a deep learning model diversity data-free quantification method based on a generator. The method comprises the steps of obtaining basic feature representation of a category label of target generation data; the generator maps generation features of different levels through a normalized stream, and each layer of attention mechanism adds one mapped noise data to learn the relationship between different categories of labels to generate synthetic data; inputting the synthetic data into a full-precision model and a quantitative model at the same time, and extracting different levels of recognition features from the full-precision model; the multi-layer feature mixer obtains identification features of the same type of synthetic data at different levels and carries out linear fusion, and a fusion result is used for initializing basic feature representation; and aligning each level of the quantitative model and each level of the full-precision model, and calibrating the quantitative model by using the distribution difference output by the full-precision model and the quantitative model. According to the method, the diversity quantification requirement of the data-free quantification model is met, and the effective improvement of the model quantification effect is ensured.

The invention discloses a data quantification method, which comprises the steps of operating an artificial intelligence AI model, acquiring data generated in the operation process of the AI model, the data comprising a plurality of groups, determining the abnormal degree of the data in the first group in the plurality of groups, the abnormal degree indicating the difference degree of the size of the data in the first group, and determining the difference degree of the size of the data in the second group in the plurality of groups. And determining parameters for performing exception suppression on the data in the first group according to the exception degree, suppressing the data in the first group according to the parameters for exception suppression, and quantifying the suppressed data in the first group. According to the method, the abnormal degrees of different groups are determined respectively, and the appropriate abnormal suppression parameters are selected for the different groups based on the abnormal degrees to perform abnormal value suppression, so that adaptive suppression of data in the different groups is realized, the situation of insufficient suppression or excessive suppression can be avoided, and the quantization precision of the AI model is improved.

The invention provides an impact deformation feature extraction method based on high-speed camera shooting and image processing, and the method comprises the steps: collecting the transient deformation data of the edge package of a remote controller housing under impact through high-speed camera shooting, recording the change speed of a strain rate, obtaining the material characteristics, the size of a package structure, and the geometric discontinuous features of an opening and closing part, generating continuous frame sequence images; performing strain field data quantification processing on the third visualization result, extracting strain field characteristic data, and judging influence weights of stress wave reflection on strain fields of different functional areas of the shell in combination with a falling height and an impact angle of the remote controller; and improving edge packaging design parameters according to the deformation buffer effect and discontinuous quantitative indexes at the opening and closing positions, verifying the improvement amplitude of the deformation buffer effect, and obtaining strain field distribution data of the optimized shell.

The invention relates to the technical field of intelligent visual detection, in particular to visual field detection equipment, system and method based on extended reality (XR), and the method comprises the steps that an image feature extraction module collects image data in real time, extracts features and eliminates optical distortion; the fixation deviation correction module analyzes eye movement data in real time, quantifies fixation deviation and triggers visual prompt and correction; the adaptive stimulation generation module dynamically encrypts to generate a test point image and adaptively adjusts the test point image according to visual field analysis and test requirements; the dynamic adjustment image module predicts and optimizes a test point presentation sequence according to image features and fixation points, automatically encrypts a high-probability defect area and quickly jumps to a key screening area; and the visual field defect identification module extracts image features, calculates the defect probability of each region in real time by using a Bayesian network in combination with subject response data, and locates a high-risk region. The system provides an efficient and accurate visual field detection scheme through the synergistic effect of multiple modules.

The invention relates to the technical field of production quality control, and discloses a production data hierarchical processing and quality tracing system. The system comprises a production data acquisition module, a data hierarchical processing module, a quality fluctuation index module, a multi-dimensional quality difference analysis module and an abnormal source positioning module. The production data acquisition module acquires a real-time parameter data sequence of each production link; the data layering processing module is used for layering the data into a plurality of data hierarchies; the quality fluctuation index module obtains a quality fluctuation index based on the fluctuation condition of the data hierarchy; the multi-dimensional quality difference analysis module analyzes the multi-dimensional quality difference based on the index; and the abnormal source positioning module positions a production abnormal source based on the multi-dimensional quality difference. According to the system, through hierarchical data processing, quality fluctuation quantification and multi-dimensional difference analysis, accurate positioning of a production abnormal source is realized, and systematicness and accuracy of production data processing and quality tracing are improved.

The invention relates to the technical field of image processing, in particular to a damper valve system adjustment method based on an artificial intelligence technology, a product, equipment and a storage medium. The method comprises the steps that parameter combinations of all valve systems of the shock absorber are collected; performing data quantization on the parameter combination to obtain a feature combination; and inputting the feature combination into a neural network model to obtain a change curve of the damping force along with the piston speed predicted by the neural network. According to the method, the external characteristic functions of the passive hydraulic shock absorber under different valve system structures are predicted through the artificial intelligence algorithm, and the number of tests in the chassis adjustment process is reduced.

The invention discloses an intelligent diversion and recycling decision support system for rural domestic sewage, and particularly relates to the technical field of sewage treatment.The intelligent diversion and recycling decision support system comprises a sensing acquisition module, an initial judgment identification module, a data quantification module, a path prediction module and a path execution module, the initial judgment identification module judges whether discharge abnormity exists or not and generates a quantization signal, the data quantization module calculates a resource potential index and a processing risk index, the path prediction module outputs a prediction path value based on the indexes, and the path execution module matches a shunting operation type according to the path value and calls a corresponding sewage processing path; according to the method, a closed-loop control mechanism from data to execution is constructed through structured acquisition of water quality data, abnormal triggering of a quantification process and double-index driving path prediction, dynamic and accurate scheduling of a sewage treatment path is realized, and the data utilization efficiency and the responsiveness and intelligence of a treatment decision are improved.

The invention provides a data quantization circuit based on a lookup table, a data processing method, a processor, equipment and a medium. The data quantization circuit includes: a data block pre-processing unit configured to determine target input data having a maximum exponent portion in an input data block; the index alignment unit is configured to align the index parts of the multiple pieces of input data to index values according to the index values of the index parts of the target input data, and multiple pieces of intermediate data are obtained; the index generation unit is configured to extract mantissa parts of a plurality of pieces of intermediate data and combine the mantissa parts with N scaling factor mantissa part candidates to obtain N first indexes corresponding to each piece of input data; and the output coding unit is configured to access the lookup table according to the N first indexes corresponding to each piece of input data, and determine a target scaling factor and a quantization result of the input data block. According to the method, the optimal scaling factor is dynamically selected through the lookup table, high quantization precision is ensured, implementation cost and delay are reduced, and hardware efficiency is improved.

The invention relates to the technical field of data processing compression and machine learning, and discloses a non-stationary data quantitative processing method based on dynamic volatility perception, which comprises the following steps of: continuously tracking real-time volatility statistical measurement of a photovoltaic data stream by adopting a Welford algorithm, and parallelly calculating a first-order differential sequence of the data stream to capture time dynamic characteristics; constructing a non-uniform quantizer with learnable parameters, setting a non-uniform input threshold by using a quantile initialization strategy, and mapping a non-uniform interval into a fixed and uniform integer output in combination with a non-uniform-to-uniform quantization mechanism; in order to guide learning of quantization parameters, a dynamic composite loss function based on volatility adaptive Beta divergence and time gradient consistency is constructed, the function comprises a Beta divergence distribution loss item dynamically controlled by real-time volatility, and the method can adapt to non-stationary characteristics of energy data such as photovoltaic power and the like. The robustness is remarkably improved while the high signal-to-noise ratio is guaranteed, and the method is suitable for edge computing scenes with limited resources.

The invention discloses a numerical control machine tool cutting force intelligent regulation and control method and system based on data analysis, and relates to the field of numerical control machine tool machining control, and the method comprises the steps of S1, data collection, S2, state evaluation model construction, S3, tool and personnel associated monitoring, and S4, hierarchical regulation and control and response. Related data in the cutting process of the numerical control machine tool are collected through various sensors; quantifying the overall state of the cutting process based on the collected data; monitoring the cutter state in real time, and associating the cutter state with personnel operation in combination with the operation habit and skill level data of an operator; rapid cutting force state judgment and response are achieved through data analysis, and a multi-stage regulation and control triggering mechanism is set; according to the method, through multi-dimensional monitoring and intelligent response, a management closed loop of collection-evaluation-monitoring-regulation is formed, an innovative solution is provided for efficient and stable machining of the numerical control machine tool, and the machining quality and efficiency are effectively improved.

The invention provides an information visualization method and system, a medium, equipment and a program product, and relates to the technical field of information processing, and the method comprises the steps: obtaining heterogeneous input information; performing category detection on the heterogeneous input information by utilizing the classification model, and determining a category corresponding to the heterogeneous input information; calculating entity relevancy, information relevancy and information weight corresponding to the heterogeneous input information according to the category; inputting structured label information corresponding to the heterogeneous input information according to the entity relevancy, the information relevancy and the information weight; constructing an atlas data structure corresponding to the structured label information; and mapping the map data structure into visual elements to obtain visual information. According to the method, heterogeneous input information is converted into visual information, so that leap-type improvement from bottom-layer data quantification to top-layer cognitive perception is realized.

A method of coding video data, the method comprising: obtaining input data, wherein the input data includes one or more of predicted video data, reconstructed video data, quantization parameter data, boundary strength data, or prediction mode data; converting the input data from an input domain to a transform domain to generate converted video data; applying a neural network (NN)-based in-loop filter (ILF) to the converted video data to generate filtered video data; and converting the filtered video data from the transform domain to the input domain.

The invention discloses a new energy station primary frequency modulation capability evaluation method considering a power grid structure and operation characteristics. The method comprises the following steps: collecting data; the data comprises complex network data of a power system, historical operation data of the system, virtual inertia control parameters of a wind generating set, droop control parameters of a photovoltaic generating set and primary frequency modulation control performance data; quantifying the primary frequency modulation comprehensive capability of the new energy station according to the collected data; and training the new energy station primary frequency modulation capability evaluation model and obtaining the new energy station primary frequency modulation capability evaluation model with the highest prediction accuracy. According to the method, the new energy station primary frequency modulation capability evaluation model is established based on the XGBoost algorithm, compared with a traditional time domain simulation method, when the method is applied to different new energy stations, modeling does not need to be conducted again, and the method has obvious advantages in the aspects of universality and calculation efficiency.

The invention discloses a tool wear monitoring method for a high-speed drilling machine, and particularly relates to the technical field of turning and boring tool wear monitoring, which comprises the following steps: deploying multiple types of sensors to acquire tool wear data, preprocessing the acquired data, generating multi-dimensional wear parameter feature vectors, and fusing the feature vectors. And generating a tool wear feature vector set, inputting the feature vector set into a pre-trained tool wear prediction model, outputting a real-time wear quantized value of the tool, comparing the wear quantized value with a preset threshold value, and generating tool wear state information and a control instruction. According to the method, the problems of single tool wear monitoring dimension, insufficient data quantization capability, low intelligent degree and lack of closed-loop control are solved by collecting the tool wear data, preprocessing the data, generating the tool wear feature vector set, outputting the tool real-time wear quantization value and generating the tool wear state information and the control instruction.

In a data storage device with a multi-level-cell memory, bits of data are stored in different levels of each of the memory cells. In some situations, a quantized version of the stored data (e.g., the most-significant bits) may be requested. Responding to such a request can involve reading all levels of the memory cells to retrieve a full version of the data and then selectively providing only the quantized version. To improve performance, the data is stored in the memory in an interleaved manner in which the quantized version of the data is stored in the same level(s) instead of being spread across all levels of the memory cells. That way, when the quantized version of the data is later requested, only the relevant level(s) are sensed, thereby avoiding the time and resources needed to read memory level(s) that do not store the quantized data.

Provided is a data quantization method, comprising: running an artificial intelligence (AI) model, and obtaining data generated in the operation process of the AI model, wherein the data comprises a plurality of groups; determining an abnormality degree of data in a first group among the plurality of groups, wherein the abnormality degree indicates the difference degree of the size of the data in the first group; on the basis of the abnormality degree, determining parameters for performing abnormality suppression on the data in the first group; on the basis of the parameters for the abnormality suppression, suppressing the data in the first group; and quantizing the suppressed data in the first group. By respectively determining the abnormality degrees of different groups, and selecting for different groups appropriate parameters for abnormality suppression on the basis of the abnormality degrees so as to suppress abnormality values, adaptive suppression of data in different groups is achieved, thereby avoiding the situation of insufficient suppression or excessive suppression, and improving the quantization precision of the AI model.

The invention relates to the field of electronic competition and data processing, and discloses a competition dynamic priority grouping and intelligent arrangement system, which comprises an event twinborn body construction module used for aggregating multi-modal data in real time and constructing a competition panoramic digital mirror image; the dynamic narrative position potential field generation module is used for quantifying and modeling multi-dimensional narrative values of all potential games in the future, such as quick cold, perennial enemy and the like, on the basis of event twinborn body data; the intelligent arrangement decision-making module takes the competition state and the narrative position potential field as input, and outputs an optimal arrangement scheme by taking maximization of competition fairness, audience participation degree and narrative value as targets through a reinforcement learning algorithm; and the narrative intervention module is used for identifying narrative opportunity points by analyzing the potential field gradient and generating an executable operation strategy to actively improve the competition attraction. Through the closed-loop design of "intervention-feedback-enhancement", the system can be continuously self-optimized, and the ornamental value and commercial value of the competition are significantly improved.

The invention relates to a processing device and a quantization method for quantization and inverse quantization of numeric data. In one or more aspects, a processing apparatus for numeric data quantization includes processing circuitry to determine a maximum exponent from a set of exponents of a set of digit representations of a set of digits, obtain a set of scaled exponents based on the maximum exponent, and quantize the scaled exponents based on the set of scaled exponents. And one of (i) obtaining a set of quantized significant numbers based on the set of digit representations and a set of mantissas of the set of scaled exponents, or (ii) obtaining a set of quantized mantissas based on the set of mantissas. The processing circuitry is configured to output a set of quantized digit representations of the set of digits based on the set of quantized significant numbers or based on the set of quantized mantissas and the set of scaled exponents; and outputting an offset exponential scaling factor based on the maximum exponent.

A computing system for performing runtime data handling optimization for generative models is provided. The computing system comprises at least one processor and memory comprising a first memory and a second memory, wherein the memory stores instructions that, when executed by the at least one processor, cause the at least one processor to execute a generative model. The computing system computes a first value matrix entry based upon the processing of an input to the generative model. The first value matrix entry is stored in a first memory wherein a first group of value matrix entries is identified. The computing system executes data quantization on the first group of value matrix entries which results in a first quantized value matrix. The first quantized value matrix is added to a second memory where it can be used during generation of the generative model. When the first group of value matrix entries is less than a group size parameter, data padding matrix values are generated and used during execution of the data quantization.

The invention relates to the technical field of interactive digital sand tables, in particular to a digital sand table interaction method and system, and the method comprises the following steps: obtaining user eyeball micro-motion trajectory data in a digital sand table display area, recording the coordinates of a sand table target point, monitoring the sight drift amplitude and direction change, calculating the sight position change amplitude in unit time, and calculating the visual position change amplitude. And establishing a sight stability change rate. According to the method, the sight stability change rate is calculated, the sight movement trend is accurately recognized, short-time jitter mistaken touch is avoided, the accuracy of interactive target recognition is improved, periodic jitter is judged based on micro-motion range change, target locking is stabilized, and accurate capture of interactive intentions is ensured. A trigger area is dynamically adjusted and highlighted, the attention concentration degree and interaction efficiency are improved, the target locking state is adjusted in real time, interaction continuity and intelligence are enhanced, interaction data are recorded, user operation habits are quantified, the misoperation probability is reduced, an instruction with the highest interaction frequency is preferentially executed, and accuracy and smoothness are improved.

According to the confidence coefficient determination method, the computer storage medium and the terminal, reference true value data is obtained based on data collected by multiple sets of multi-source fusion integrated navigation positioning terminals, and a premise is provided for reliable quantification of the confidence coefficient; reference true value data is used as comparison data, quantitative reference data are selected from data collected by multiple sets of satellite navigation positioning terminals, and reliable sample data used for obtaining a confidence coefficient calculation function is obtained; quantizing a predetermined index to obtain a quantized score, and constructing a confidence calculation function of an initial quantized confidence based on the quantized score and the weighting coefficient; according to the method, weighting coefficients are adjusted through iteration, obtained confidence verification values of different scenes are all in corresponding confidence target intervals, a confidence calculation function with calculated confidence conforming to objective reality is obtained, the confidence of actual positioning data is quantified through the obtained confidence calculation function, and the accuracy of the actual positioning data is improved. And the confidence degree of the actual positioning result is analyzed.