28 results about "Data truncation" patented technology

The invention discloses an emotion recognition method based on an attention mechanism and a capsule network. The emotion recognition method specifically comprises the following steps: inputting multi-channel original EEG signals; performing data truncation, data filling and data standardization on the original EEG signal, unifying the format, and generating a preprocessed EEG signal; extracting space-time joint features from the preprocessed EEG signals by using a three-dimensional convolutional neural network; the construction of the graph structure comprises modeling a complex relationship between channels; introducing a graph attention mechanism to strengthen information expression of important channels and relationships; and high-dimensional vector modeling and final classification of the features are carried out through a capsule network, and capture of deeper space structure information is completed.

The invention discloses an offset scanning CBCT image reconstruction method. The method comprises the following steps: virtual detector conversion of projection data; carrying out geometric weighting on the data; carrying out data continuation and filtering; parker weighting is carried out on the data; and reconstructing bias back projection. According to the invention, aiming at the scanning mode of radiation source and detector bias, the FDK algorithm cannot be directly used for reconstruction, the virtual detector is introduced to convert the scanning mode into the scanning mode of detector bias, the data is subjected to continuation processing to solve the problem of data truncation, and the filtered data is subjected to Parker weighting aiming at the repeated scanning area to solve the problem of bias artifacts. And finally, carrying out three-dimensional reconstruction by adopting an FDK reconstruction method. The reconstruction method is high in speed, and the obtained cross-sectional image is high in definition.

The application relates to the technical field of data processing, and discloses a network public opinion hotspot evaluation system based on sentiment analysis, which comprises the following modules: a data analysis module, which is used for receiving an interactive flow extraction identifier and text and calculating a basic sentiment polarity value; a window control module, which outputs a current time window span based on a topological entropy gradient of a previous period; a topological layering module, which constructs a local directed graph to calculate a global topological entropy, and takes a shortest path hop count as a topological depth; a polarization rate calculation module, which allocates and calculates a variance of the polarity value according to the topological depth and outputs a sentiment polarization rate; and a comprehensive evaluation module, which combines a word frequency, the sentiment polarization rate and an activation function to output a comprehensive hotness evaluation value. The application adaptively adjusts data truncation according to network public opinion fluctuation, prevents feature dilution through spatial layering sentiment analysis, and automatically peels off abnormal concurrent flow based on a topological base, so that the hotspot evaluation is real and objective.

The present application relates to the field of spaceflight TT&C technology, and specifically discloses a method and system for implementing an extraterrestrial object blocking TT&C plan under deep space large time delay conditions. The present application determines extraterrestrial object blocking time windows and ground available tracking stations by using precise orbit data of a probe, calculates station prediction containing data truncation and one-way light time of uplink and downlink signal transmission, sets uplink tracking events in advance and downlink tracking events in arrears according to the one-way light time, merges adjacent TT&C arcs with a time interval less than a preset threshold into single-station continuous tracking arcs to set task preparation events, and decouples uplink and downlink under deep space TT&C mode, controls transmission of uplink carriers and remote control instructions according to the one-way light time, thereby solving the technical problems of difficulty in space-earth cooperation and poor timeliness of emergency retransmission during extraterrestrial object blocking under deep space large time delay conditions, and improving the reliability and timeliness of implementation of deep space exploration mission TT&C plans.

The invention discloses a differential dynamic microscopic image processing method based on a root-mean-square error. The differential dynamic microscopic image processing method comprises the following steps: a sequence image acquisition step: acquiring a time sequence microscopic image; a sequence image processing step: obtaining a difference image data set; performing one-dimensional Fourier transform in two directions on the differential image data set in sequence to obtain two-dimensional transform result data, and calculating to obtain Fourier amplitude spectrum image data; an obtained data processing step: carrying out dimension reduction processing on the Fourier amplitude spectrum image data; introducing structure function fitting, calculating a root-mean-square error between a fitting curve of the data after dimension reduction processing and the Fourier amplitude spectrum image data, and selecting a data truncation point according to the root-mean-square error; and an intermediate scattering function is introduced, and dynamic information of Brown particles is obtained based on the relationship between the data after selection of the truncation point and the diffusion coefficient. According to the method, the data truncation point is determined through the root-mean-square error, so that the processing time is remarkably shortened while the measurement precision is ensured.

The invention provides a marketing suggestion generation method, device and equipment for a target community and a storage medium, and the method comprises the steps: selecting a full-field user feature data stream of a target user group, which accords with a preset rule, from a user portrait system of the target community, reading a data compression operator combination corresponding to a marketing report type from an operator configuration library, each operator defines a specific data extraction and compression rule. And sequentially executing each data compression operator, performing multi-dimensional analysis and information extraction on the user feature data, generating a compressed feature description text fragment, and compressing the original mass data of millions of tokens to the scale of hundreds of tokens. According to a predefined template structure, all feature description text segments are combined in order to form a user portrait abstract, the user portrait abstract is embedded into a complete cue word template based on a feed-shot cue word construction method, it is ensured that LLM can generate a precise marketing strategy based on complete and refined user information, and the problems of data truncation and attention loss are effectively avoided.

The invention is suitable for the technical field of power spot market clearing, and provides a power delivery in-province power spot market clearing method, and the method comprises the steps: carrying out the calculation of the delivery superposed power of a target unit in a current transaction period based on the obtained delivery power data and technical parameter data; generating a delivery newly-added curve segment based on the delivery superposed power and the acquired market price lower limit; performing data translation processing on each section of the original declaration curve data based on the outgoing superposed power, and performing data truncation processing on the declaration curve data after data translation processing to obtain an effective curve section after data truncation processing; performing data splicing on the delivery newly-added curve segment and the effective curve segment to generate a reconstructed declaration curve of the target unit; based on the reconstructed declaration curve, clearing the current transaction period in the power spot market in the province; the accuracy of clearing calculation input data is ensured, and the reliability of clearing price signals in the provincial spot market is improved.

This invention relates to the field of electricity meter monitoring, and more particularly to an online monitoring system and method for smart electricity meters. The system includes: periodically determining the link transmission coefficient of each electricity monitoring set based on the link throughput fluctuation index and the link balance index; determining whether each electricity monitoring set meets the transmission optimization conditions based on the link transmission coefficient; under the transmission optimization conditions, determining the data mining strategy to be executed for each target monitoring device based on the data acquisition status; when performing time series analysis, determining whether to perform data truncation compensation based on the interval between each time series truncation segment, and determining the truncation compensation method based on the electricity load coefficient of the truncation coverage area; under the condition that the set analysis is completed, determining whether to perform transmission link optimization for the target optimized set based on the proportion of key equipment and the reference equipment characteristic index, and determining the allocation transmission strategy based on the demand allocation difference value. This invention improves the data processing efficiency in the monitoring process of smart electricity meters.

The invention relates to a method for realizing spectrum analysis processing based on an FFT (Fast Fourier Transform) and frequency sweeping combination mode. The method comprises the following steps of: performing switching control through a scanning control state machine; selecting a filter according to the frequency resolution bandwidth by extracting a low-pass filter bank; windowing is carried out on the extracted IQ data in a time domain; carrying out FFT (Fast Fourier Transform) operation; calculating the FFT data truncation length, and carrying out exchange sorting processing on the FFT data after truncation; calculating the IQ data amplitude of the FFT data after truncation sorting processing; and performing detection processing on the amplitude data. The invention also relates to a frequency spectrum analysis device for realizing the combination of FFT and frequency sweeping, a processor and a readable storage medium. According to the method and device for realizing spectrum analysis processing based on the FFT and frequency sweeping combination mode, the processor and the computer readable storage medium, under the conditions of large scanning width and high frequency resolution, the speed is higher than that of traditional frequency sweeping, the continuous scanning effect is different from that of traditional frequency sweeping, and the higher the frequency resolution is, the more obvious the effect is.

The invention discloses a building system self-adaptive control method based on data truncation and modular transfer learning, and relates to the technical field of intelligent building energy consumption optimization and control, modular modeling is performed on each component of a building system, a hybrid recurrent neural network is adopted as the architecture of a building system modular model, and the intelligent building energy consumption optimization and control is realized. The hybrid recurrent neural network comprises a plurality of recurrent neural network units, and each recurrent neural network unit corresponds to each component module in the building system; after sudden change of the building system is detected, only module parameters influenced by sudden change are unfrozen and finely adjusted, and module parameters not influenced by sudden change are kept frozen; meanwhile, data before mutation is abandoned to eliminate the influence of an old domain, and only the data after mutation is reserved to perform fine adjustment on unfrozen module parameters; in addition, in the model transfer learning process, only the prediction error output by the module influenced by sudden change is used as a loss function. The invention aims to solve the problems that the domain migration efficiency is low and old domain data interferes with new domain learning.

The invention discloses a verifiable outsourcing computing privacy protection method based on block data truncation, which adopts a three-party architecture of participants, an aggregator and a key generation center, and takes a non-collusion cloud node as a data block distribution and computing carrier. The key generation center constructs a distributed multi-trap-door public key cryptosystem based on security parameters, splits strong private keys and distributes the strong private keys to participants and aggregators. A participant splits original data into mutually exclusive blocks through binary segmented truncation and remainder operation according to the bit number of sensitive data, and distributes the blocks to different cloud nodes. And the cloud nodes complete data block addition, multiplication and equivalence verification by adopting a carry-aware mechanism to ensure that a calculation result is consistent with original data operation. And the aggregator collects a result and performs cross check, and the collaborative participant decrypts the result and then integrates the result to obtain a final weighted aggregation result. According to the method, the problem of unbalance among verifiability, privacy and practicability of the existing outsourcing computing technology can be solved.

The invention discloses a data synchronization storage space adjustment method and device, equipment and a storage medium. The method comprises the following steps: acquiring a source end storage unit of a storage unit used by a target source end and a standby end storage unit of a storage unit used by a target standby end; whether the source end storage unit and the standby end storage unit conform to preset storage unit types or not is determined, and the preset storage unit types comprise the first storage unit type and the second storage unit type. And determining a configuration item value corresponding to a matching type according to the source end storage unit, the standby end storage unit and the matching type of the preset storage unit type. According to the configuration item value and the target synchronization data requested by the target source end, the target storage space used for storing the target synchronization data in the target backup end is adjusted, so that the reliability of the data backup disaster recovery system is guaranteed, the data truncation and messy code risks in the data migration process are reduced, the data synchronization efficiency is remarkably improved, and the user experience is improved. And the data integrity and accuracy are ensured.

This invention discloses an adaptive data truncation method and system based on variance goodness-of-fit, relating to the field of natural language processing technology. The method includes acquiring multiple documents output by a re-ranking model and their corresponding relevance scores; arranging the relevance scores in descending order to obtain an ordered relevance score sequence; starting with a cluster size of 2, progressively increasing the cluster size; and classifying the ordered relevance score sequence using a natural breakpoint classification method for each cluster size. This invention performs natural breakpoint classification and variance goodness-of-fit calculation on the ordered relevance score sequence with multiple cluster sizes, then determines the optimal cluster size based on the variance goodness-of-fit gain value of adjacent cluster sizes, and adaptively determines the truncation threshold. This allows for dynamic adjustment of the filtering boundary based on the score distribution characteristics under different query scenarios, improving the adaptability of data truncation and reducing interference from low-relevance documents.

The invention relates to the technical field of seismic data processing, and particularly discloses a self-adaptive Hanning window construction method and device, and the method comprises the steps: giving an initial adjustment parameter of an initial window function, and calculating a flag value of the initial window function; setting an adjustment interval of the mark value and a correlation condition satisfied by the initial adjustment parameter and the mark value; obtaining a final adjustment parameter through iterative operation according to the adjustment interval and the association condition; and obtaining a calculation formula of the final window function according to the final adjustment parameter. The self-adaptive Hanning window constructed by the method changes more gently in starting, ending and intermediate function value, the truncation effects of data truncation, data filtering, data fusion and the like of actual data are obviously weakened, the practical requirements are better met, and the use adaptability and convenience in the practical process are improved.

This invention provides a Maxout activation function acceleration device and method based on in-memory computing devices. The device includes: a data splitting module for splitting data into integer and fractional parts; a data truncation module for splitting the integer and fractional parts into fixed-bit lengths in binary form and transmitting them to an in-memory computing device array, while simultaneously generating corresponding shift control signals and transmitting them to an accumulator module; a comparator module for comparing the array's output, obtaining the maximum value, and outputting it; an accumulator module for shifting and accumulating the comparator module's output and outputting the accumulated result according to the shift control signal; and a weight queue storage module for storing the weights corresponding to the neural network workload and updating the weights stored in the in-memory computing device array after completing one layer of computation. This invention offers advantages such as low power consumption, high energy efficiency, adjustable precision, and low computational latency.

A magnetic field speed sensor includes one or more magnetic field sensing elements to generate a magnetic field signal indicative of a magnetic field associated with a moveable target and a controller to generate a controller output signal including a plurality of output words, each including a speed pulse and a plurality of data pulses capable of having four different amplitudes. The time between the speed pulses of consecutive output words is indicative of the speed of movement of the target. Each data pulse corresponds to a data bit having a binary value represented by a transition direction of the data pulse and the amplitude of the data pulse. Features include a programmable data pulse width or an adaptively adjustable data pulse width to avoid truncation.

Systems and methods for data truncation are disclosed. A computing system may a storage medium and a processor. The processor may be configured to identify first data of a first data structure; identify a first portion of the first data and a second portion of the first data; store the first portion of the first data as a second data structure in the storage medium, and the second portion of the first data as a third data structure in the storage medium; identify a request for the first data; based on the request, retrieve from the storage medium the second data structure including the first portion of the first data; identify a value associated with the second portion; generate second data based on the first portion and the value; and perform a task based on the second data.

Methods for asynchronously determining relational data integrity using cryptographic data structures are performed by systems and devices. Changes in current tables of relational databases are reflected in associated history tables. Cryptographic hybrid blockchain ledgers are updated with transaction records, for entry changes in current and history tables, including transaction information and hash values of corresponding entry changes. Hybrid blockchain ledgers also include root hash values of Merkle trees of transaction records in current blocks, and hash values of prior blocks. A current block receipt is asynchronously generated and provided as a single hash value from which the validity states of the tables and ledger are able to be verified. Cryptographic receipts of specific transactions reflected in table entry changes are generated and provide immutable evidence of specific transaction existence for users. Ledger-enabled tables are provided for mixed database operations with ledger-disabled tables, and temporal history table database operations are enabled.

The invention provides a data truncation method based on a large model and a related device. Specifically, the nesting depth, the node path and the node value of each data node in the to-be-truncated data can be respectively determined, the semantic importance of each node path is respectively evaluated by adopting a semantic analysis model, and the data can be truncated based on the semantic importance of the node paths and the nesting depth of the data nodes. And comprehensively judging whether data truncation operation needs to be executed on the data nodes or not. Therefore, key information and non-key information in the to-be-truncated data can be recognized in combination with the hierarchical relationship of the data nodes, the nested structure and the semantic importance of the path fields, and data truncation is performed accordingly, so that the data volume can be reduced under the condition that the key information of the data is reserved, and the data quality of the truncated data can be improved.

The invention discloses a sensor digital output data compression method, and relates to the technical field of sensor chip design, and the method comprises the steps: obtaining the original digital output data of a sensor; determining a data interception bit width N and a bit width K of a scaling factor SF based on the bit width M of the original digital output data in combination with an error requirement of a digital output dynamic range of the sensor and a data volume compression requirement; intercepting N-bit data from the original digital output data to obtain effective data segments; determining the value of a scaling factor (SF) according to the position of the effective data segment in the original digital output data; and combining the effective data segment and the scaling factor SF to obtain compressed data output by the sensor. The method maintains an extremely high dynamic range while significantly reducing the amount of data by compressing the bit width of the original digital output data.

The application discloses a kind of energy storage battery management devices based on AI computing power, it is related to energy storage battery management technical field.The device obtains the relaxation voltage signal after charging cut-off under long-term shallow charge-discharge condition of lithium iron phosphate energy storage cell by signal acquisition unit, risk assessment unit is according to the signal to assess the potential abnormal evolution risk of each cell under multiple cells concurrent scene and generate high-risk cell identification, simultaneously determine timing characteristic window, and power scheduling unit adopts timing characteristic window locking distribution algorithm to adjust the AI inference resource scheduling mode from fixed polling, average distribution or shared call to window locking distribution mode, and signal processing unit executes modal decomposition under locking distribution mode to separate abnormal fluctuation component to relaxation voltage signal.The application assesses risk and locks window, and converts computing power scheduling to continuous distribution in window, avoids time series data truncation, separates abnormal component under continuous computing power and executes inference, captures hidden abnormality and early warning.

Embodiments of this specification disclose secure multi-party computation for privacy protection. In an implementation, a method includes obtaining a fragment of first gradient data and a fragment of noise data, where the first gradient data is gradient data of a loss function. Based on the fragment of first gradient data by performing secure multi-party computation with another participant, obtaining a fragment of second gradient data, where the second gradient data is gradient data obtained after the first gradient data is clipped. Based on the fragment of second gradient data and the fragment of noise data, determining a fragment of third gradient data, where the third gradient data is the second gradient data with the noise data added. A fragment of a model parameter is determined based on the fragment of third gradient data.

This invention relates to the field of computer data processing technology and discloses a dynamic regularization and collaborative scheduling method for time-series information data. The method accesses asynchronous multi-source time-series data streams, initializes a regularization pool based on a global tracking identifier, and sets a basic survival time window. It scans the regularization pool to calculate the associated state feedback factor and collects execution layer load data to calculate the resource water level. When the feedback factor reaches the high readiness state locking threshold and the resource water level is below the congestion threshold, the regularization pool is locked, a dedicated thread is allocated, and a data block access descriptor is generated for reading and computation. If the regularization pool approaches the time window and the data is not fully ready, the execution state dimensionality reduction or time window compensation operation is dynamically selected based on the resource water level status. This invention avoids memory accumulation and effective data truncation problems caused by fixed time windows by establishing a joint determination mechanism of data readiness and underlying resources, thereby improving the utilization and throughput of computing resources in concurrent scenarios.

This invention relates to the field of computer data processing technology and discloses a dynamic regularization and collaborative scheduling method for time-series information data. This method accesses asynchronous multi-source time-series data streams, initializes a regularization pool based on a global tracking identifier, and sets a basic survival time window. It scans the regularization pool to calculate the associated state feedback factor and collects execution layer load data to calculate the resource water level. When the feedback factor reaches the high readiness state locking threshold and the resource water level is below the congestion threshold, the regularization pool is locked, a dedicated thread is allocated, and a data block access descriptor is generated for reading and computation. If the regularization pool approaches the time window and the data is not fully ready, the execution state dimensionality reduction or time window compensation operation is dynamically selected based on the resource water level status. This invention avoids memory accumulation and effective data truncation problems caused by fixed time windows by establishing a joint determination mechanism of data readiness and underlying resources, thereby improving the utilization and throughput of computing resources in concurrent scenarios.

This invention relates to the field of aerospace telemetry and control technology, specifically disclosing a method and system for implementing telemetry and control plans under deep-space conditions with large time delays due to extraterrestrial celestial body obstruction. This invention determines the time window for extraterrestrial celestial body obstruction and the available ground tracking stations by utilizing precise orbital data from the probe. It calculates the station forecast including data truncation and the one-way optical travel time of uplink and downlink signal transmission. Based on the one-way optical travel time, it pre-sets uplink tracking events and delays downlink tracking events. It merges adjacent telemetry and control arcs with time intervals less than a preset threshold into a single-station continuous tracking arc to set mission preparation events. In deep-space telemetry and control mode, it decouples the uplink and downlink, and controls the transmission of uplink carriers and remote control commands based on the one-way optical travel time. This solves the technical problems of difficult space-ground coordination and poor timeliness of emergency resending during extraterrestrial celestial body obstruction under deep-space conditions with large time delays, improving the reliability and timeliness of deep-space exploration mission telemetry and control plans.

The application discloses a large model hardware deployment processing method, a chip and an electronic device. The method comprises the following steps: acquiring the quantization sensitivity of each operation unit of a large model during deployment and running on a hardware device; identifying a quantization sensitive target which meets the set requirement for the influence on the hardware calculation error according to the quantization sensitivity; obtaining a first quantization coefficient for controlling data truncation in a manner independent of training based on the quantization sensitive target, so as to minimize the quantization output error of the hardware device; when the optimization result of improving the quantization precision based on the first quantization coefficient does not meet the requirement or the quantization precision needs to reach the set standard, performing layer-by-layer optimization based on the quantization sensitive target by using a training method to obtain a second quantization coefficient for controlling data truncation; and generating a large model hardware deployment processing file suitable for the hardware device to perform reasoning based on the first quantization coefficient or the second quantization coefficient. The application can improve the deployment efficiency of the large model.

Systems and methods for data truncation are disclosed. A computing system may include a storage medium and a processor. The processor may be configured to: identify first data of a first data structure; identifying a first portion of the first data and a second portion of the first data; storing a first portion of the first data as a second data structure in the storage medium, and storing a second portion of the first data as a third data structure in the storage medium; identifying a request for first data; retrieving, based on the request, a second data structure including a first portion of the first data from a storage medium; identifying a value associated with the second portion; generating second data based on the first portion and the value; and execute the task based on the second data.

The present application discloses a color correction method, apparatus, device and storage medium, belonging to the field of image processing technology. The method includes obtaining overflow data and truncated data of a first image converted from an RGB color space to a uniform color space, the overflow data being used to represent data actually acquired by an image acquisition device and unable to be displayed by a display device, and the truncated data being used to represent data that can be displayed by the display device; performing data fusion processing on the overflow data and the truncated data to obtain first color data of the first image in the uniform color space, the first color data including first brightness data, first hue data and first chromaticity data; generating a second image in the RGB color space based on the first brightness data, the first hue data and the second chromaticity data, wherein the second image is an image of the first image after color correction, and the second chromaticity data is determined by relative chromaticity data corresponding to the first chromaticity data.