202 results about "Sample sequence" patented technology

The invention relates to the technical field of urban rail transit station-city collaborative planning, in particular to an ontology-based station-city collaborative data integration and planning prediction method, which comprises the following steps of: obtaining rail transit station passenger flow data, resident travel behavior data and station periphery built environment index data; forming a space-time sample sequence according to the unified space-time granularity of the site walking service area; constructing an urban rail transit station-city cooperation ontology, and carrying out semantic annotation and semantic fusion on the space-time sample sequence to generate a feature sequence; inputting the feature sequence into a multi-task space-time diagram convolutional neural network prediction model to output a passenger flow prediction result and establish an environment index prediction result; and calculating a feature contribution degree based on a Shapley additive interpretation value, optimizing a background sample set by using a genetic algorithm to determine a key action element set, outputting a planning index threshold and an intervention measure parameter, and realizing an interpretable station-city collaborative prediction and planning decision closed loop.

The invention discloses a PCB component layout method, device and equipment based on a multi-modal large model, and relates to the technical field of electronic design automation, the method comprises the following steps: carrying out quality screening on a plurality of PCB design files which have been laid out to obtain a qualified design file set; constructing a serialized training sample set based on the qualified design file set; each training sample sequence comprises a plurality of training samples arranged according to a layout logic sequence; each training sample comprises multi-modal data representing a current layout state and a real position coordinate of a next component to be laid out; based on the serialized training sample set, training the open-source multi-modal large model to obtain a PCB component layout model; and iteratively generating a layout file of the target PCB based on the netlist file of the target PCB and the PCB component layout model. The method is high in generalization capability and low in cost based on the PCB component layout model trained by the layout data, and can realize rapid layout of components of a new PCB.

The invention relates to the technical field of artificial intelligence application, and discloses an enzyme EC number prediction method, and the method comprises the steps: obtaining the sample sequence characteristics of a to-be-predicted sample containing a substrate SMILES sequence and a product SMILES sequence through a target BERT model; constructing a molecular object and feature coding based on atom mapping, atom truncation and sequence analysis, constructing a reaction graph of a to-be-predicted sample, inputting the reaction graph into a target graph isomorphic neural network, and constructing molecular graph features of the to-be-predicted sample based on a recursive neighborhood aggregation mechanism; and fusing the sample sequence features of the to-be-predicted sample with the molecular map features by using a bidirectional cross attention mechanism to obtain multi-modal features, inputting the multi-modal features into the multi-layer perceptron, and obtaining the prediction probability of the enzyme EC number of the to-be-predicted sample. According to the method, efficient and accurate end-to-end prediction of enzyme EC numbering is realized through the multi-dimensional chemical spatial characteristics of the collaborative modeling reaction.

The invention discloses an intelligent acquisition risk identification system based on deep learning, and the system comprises a time sequence sample construction module which is used for building a time sequence sample sequence; the hierarchical attention structural feature coding module is used for carrying out structural feature coding on the sequential sample sequence by utilizing a hierarchical attention network; the time sequence feature extraction module is used for inputting the structural feature vector into an ETSform model to perform time sequence feature extraction; the hierarchical time sequence collaborative attention adaptive fusion module is used for performing bidirectional attention interaction and dynamically generating a hierarchical weight and a time sequence weight through a meta-learning controller; the improved CatBoost risk identification module is used for outputting a risk score and a risk type label; and the system fusion module is used for summarizing and merging the risk scores and the risk type labels. The method and the device are suitable for merchant transaction risk identification in an acquiring business scene.

The invention discloses a radio frequency chip phase deviation calibration method based on multichannel sampling data analysis, and relates to the technical field of radio frequency signal processing and multichannel data synchronous calibration. Processing the sampling sequence of each channel by adopting a phase estimation algorithm to calculate a phase value to obtain initial phase distribution; according to the initial phase distribution, if the number of detection channels is increased or decreased, the reference phase is recalculated by fusing the time sequence relation adjustment information to obtain an updated reference phase; acquiring an updated reference phase, and extracting a reference signal from the residual channels and compensating a missing part through an adaptive filtering algorithm to obtain compensated phase distribution if a failure channel signal is lost according to partial channel failure judgment; the radio frequency chip phase deviation calibration method based on multichannel sampling data analysis has long-term self-calibration and self-adaption capabilities, and the robustness, reliability and signal processing precision of a system are remarkably improved.

The invention discloses a wind power gear box variable speed fault diagnosis method based on LMSRCT and a medium, and belongs to the field of wind power fault detection.The method comprises the following steps that a vibration acceleration sensor is installed on a wind power gear box, an original vibration signal x (t) of the gear box in the running state is collected at the sampling frequency f s, and meanwhile a rotating speed pulse signal is collected; obtaining accurate rotation frequency f < r > (t) and shaft rotation angle information theta (t); preprocessing the collected original vibration signal x (t) to obtain x pre (t); the x pre (t) is converted to an angle domain through an LMSRCT algorithm, and a one-dimensional angle domain sequence signal s (theta) is generated; the s (theta) is segmented into a sample sequence with a fixed length L for embedded encoding, the sample sequence is input to a four-layer Transform encoder, and probability distribution of different fault types is output; and taking the fault type corresponding to the maximum probability value as a final diagnosis result. According to the method, the LMSRCT algorithm is adopted to convert the vibration signal from the time domain to the angle domain, the problem of diagnosis failure caused by spectrum aliasing is solved, and the accuracy of fault diagnosis is improved by combining the LMSRCT algorithm with a Transform model.

The invention provides an industrial product quality prediction method based on geometry preserving cross-scale difference, and relates to the technical field of industrial product quality prediction.The method comprises the steps that collected time sequence data of industrial process variables are preprocessed, a sample sequence is constructed through a sliding window, and the sample sequence is divided into a training set, a verification set and a test set according to the time sequence; the method comprises the following steps: constructing a double-branch coding architecture to independently process trend features and differential features; a geometric perception attention mechanism is introduced into each branch encoder, it is ensured that hidden layer representation and output target space keep geometric consistency, and the stability and interpretability of the model are enhanced; and deep interaction and adaptive fusion of double-branch information are further realized by adopting cross-scale cross attention, so that the comprehensive modeling capability of long-term trend and short-term dynamic in the industrial process is remarkably improved.

The invention discloses a data processing method and device, equipment and a medium. Comprising the steps that a training sample set is obtained, the training sample set comprises a plurality of sample pairs, each sample pair comprises an image sample and a text sample, and through a to-be-trained model, prediction is carried out based on the image samples and the text samples to obtain an output lexical element sample sequence, determining a first target lexical element with relatively high confidence and a second target lexical element with relatively low confidence from a plurality of output lexical elements contained in the output lexical element sample sequence; filtering the loss value corresponding to the first target lexical element to obtain a first target loss value; performing upper limit constraint on the loss value corresponding to the second target lexical element to obtain a second target loss value; and training based on the first target loss value and the second target loss value to obtain a visual language model. According to the technical scheme, the reliability of data processing in a visual language model scene is improved.

The invention relates to the technical field of precision shell contour metering, and discloses an online metering method for a precision shell contour, and the method comprises the steps: obtaining an original sampling point cloud and a feature point group of a to-be-measured workpiece in a conveying state; resolving a real-time pose matrix of the to-be-measured workpiece relative to a preset measurement reference by using a spatial topology constraint relationship of the feature point groups; carrying out differential processing on a real-time pose matrix in the sampling sequence, and synthesizing an instantaneous motion vector corresponding to a sampling moment; in combination with a signal response time delay constant of the sampling system, phase compensation is carried out on a motion displacement deviator induced by signal conversion delay, and a corrected sampling point cloud is generated; according to the method, the geometric distortion of the point cloud generated by response delay of the sensor is solved, the problem of instantaneous drift of the measurement reference in an unsteady state conveying state is solved, and the contour reduction precision in a complex working condition is improved.

The invention discloses a large model security evaluation method based on multi-dimensional adversarial attack, and the method comprises the steps: generating an adversarial sample sequence through constructing multi-dimensional attack scene description and combining the characteristics of data availability damage and system integrity violation, optimizing sample parameters in a black box attack mode, and precisely positioning a weak link of a model. And meanwhile, based on a risk quantized value sequence and a safety portrait mechanism, reinforcing demand data is extracted and a protection path is generated through reverse optimization, and finally, the protection capability of the model is remarkably improved. Through a closed-loop mechanism of scene generation, sample optimization and risk assessment, a complex attack environment is effectively dealt with, and the safety and stability of the system are guaranteed.

This invention discloses a method and system for real-time analysis of transformer operating status oriented towards edge computing, specifically relating to the field of power equipment condition monitoring and edge computing data processing technology. The method includes acquiring load values, oil temperature values, winding temperature values, cooling status values, and status monitoring values ​​collected from the transformer site. It generates three types of change symbols (increase, decrease, and remain unchanged) for each type of value according to the sampling order, and binds each change symbol to its corresponding sampling sequence number, outputting ordered transformer operating data. By converting the load values, oil temperature values, winding temperature values, cooling status values, and status monitoring values ​​collected from the transformer site into change symbols with sampling sequence numbers, it identifies load change segments or cooling change points and extracts the current temperature rise response segment. Then, it compares the current temperature rise response segment with existing temperature rise response segments in response order, and generates transformer operating status analysis results based on the order differences formed by oil temperature reversal, winding temperature changes, and status monitoring value changes.

The invention discloses a tunnel illumination control method and device, equipment, a storage medium and a program product, and relates to the technical field of tunnel illumination control, and the method comprises the steps: synchronously obtaining the operation state time sequence data and distribution position data of brightness sensors disposed on the whole line of a tunnel in a monitoring time window; wherein the operation state time sequence data comprises an original illumination sampling sequence and a real-time temperature sampling sequence of each brightness sensor; inputting the operation state time sequence data and the distribution position data into an illumination value calibration model to obtain a calibration illumination value of each brightness sensor; and performing tunnel lighting control based on the calibrated illuminance value. According to the invention, accurate control of tunnel illumination is realized.

This invention discloses a non-destructive testing method for railway slope anchor cables based on intelligent sensors, comprising: acquiring multi-source time-series data and preprocessing it to form standardized input data; determining event times and event windows, and generating event marker sequences; constructing features to obtain input sample sequences; constructing an improved liquid neural network to obtain diagnostic results; performing criterion calculation and consistency verification to generate early warning trigger conditions; employing adaptive large-scale neighborhood search to obtain an updated parameter set; and performing online updates and risk classification early warnings to form evaluation results and early warning information. This invention combines an improved liquid neural network with adaptive large-scale neighborhood search parameter optimization to achieve stable and repeatable non-destructive testing and risk classification early warning under complex railway operating conditions.

The invention discloses a flywheel fault diagnosis method based on a double-flow multi-scale Transform, and the method comprises the steps: obtaining a vibration acceleration signal of a flywheel system, and segmenting the vibration acceleration signal into a sample sequence through a sliding window technology; performing standardization processing on the sample sequence to obtain one-dimensional time domain data, and performing time-frequency conversion on the sample sequence to obtain a two-dimensional time-frequency map; constructing a double-flow multi-scale Transform diagnosis model, wherein the double-flow multi-scale Transform diagnosis model comprises a time domain branch network, a frequency domain branch network and a cross attention fusion module; inputting the one-dimensional time-domain data into a time-domain branch network to extract time-domain features, and inputting the two-dimensional time-frequency atlas into a frequency-domain branch network to extract frequency-domain features; fusing the time domain features and the frequency domain features by using a cross attention fusion module to obtain a fused fault feature vector; and classifying the fused fault feature vectors, and outputting a fault diagnosis result of the flywheel system. The method has high diagnosis accuracy under strong noise interference.

The application relates to a low-overhead sampling analysis method for a broadband wireless signal, which comprises the following steps: obtaining an original low-rate IQ sample sequence of a target broadband signal, carrying out multi-dimensional feature embedding to obtain an original feature sequence; carrying out autocorrelation analog sampling processing on the original low-rate IQ sample sequence to obtain an enhanced autocorrelation feature sequence, splicing the enhanced autocorrelation feature sequence with the original feature sequence to form a fusion enhanced feature sequence; inputting the fusion enhanced feature sequence into a spectrum sensing model based on a deep neural network for processing, predicting spectrum occupation prior information in real time, and using the spectrum occupation prior information as a signal reconstruction constraint to reconstruct the original low-rate IQ sample sequence to obtain an effective recovery signal; and using a signal analysis model based on a Transformer to carry out signal analysis on the effective recovery signal to obtain a deep analysis result. Compared with the prior art, the application has the advantages of breaking through the sampling rate limitation in the prior art and greatly simplifying a data processing procedure.

The invention provides a training method of a sequence generation model, and a sequence generation method, device and equipment, and relates to the technical field of artificial intelligence, the training method of the sequence generation model comprises the following steps: adding boundary lexical elements in an original sample sequence to obtain a first intermediate sequence; performing noise addition processing on the first intermediate sequence to generate a training input sequence; wherein the noise adding processing comprises random discarding processing and random mask processing; inputting the training input sequence into an encoder model based on a bidirectional self-attention mechanism to obtain a first insertion prediction result and a first content prediction result; and based on the first insertion prediction result, the first content prediction result and the original sample sequence, total loss is calculated to update parameters of the encoder model, and the sequence generation model is obtained. According to the invention, the sequence generation speed and the sequence generation quality can be improved.

This invention discloses an intelligent mask sealing detection system and method based on active biomimetic fit adjustment, belonging to the field of intelligent protective equipment detection technology. The system includes the following steps: collecting breathing change information and fit contact feedback information between the mask and face throughout the entire mask-wearing process, and organizing this information according to the temporal relationship between the inhalation and exhalation phases to generate a breathing fit rhythm sequence reflecting the correspondence between breathing fluctuations and fit changes. This invention uses the breathing fit rhythm as its core to dynamically analyze the fit state, identifying short-term air leakage phenomena where the time deviation of the fit response during breathing rhythm switching is synchronized with the inhalation peak, thus avoiding misjudgments of sealing. Furthermore, based on the leakage distribution pattern, it pre-adjusts the fit adjustment and sampling sequence to ensure that the detection covers key change areas, improving the reliability of mask protection under complex breathing conditions.

This invention discloses a method and system for radio frequency fingerprinting of narrowband IoT transceivers. The method involves a receiver acquiring wireless messages sent by the transceiver to be identified, obtaining in-phase / orthogonal discrete sample sequences, locating and extracting preamble sample segments, performing a short-time Fourier transform on the preamble sample segments to generate a spectrogram energy matrix, and extracting compressed time-frequency features through singular value decomposition. The main component of the carrier frequency offset is obtained based on the cross-correlation peak position shift between rising and falling chirps, and the carrier frequency offset correction component is obtained based on the phase drift of adjacent rising chirps. The compressed time-frequency features and the carrier frequency offset estimation results are concatenated into a fused feature vector, which is then input into a convolutional neural network classification model to output the device category and category confidence. This invention reduces the dependence on channel location features and decreases model input overhead, making it suitable for online identification of narrowband IoT devices.

The application discloses a coronary angiography blood vessel tracking method and system based on a template matching method, belongs to the technical field of blood vessel image processing, and comprises the following steps: acquiring any adjacent frame angiography image in an angiography sequence, acquiring a sample point sequence, and marking a to-be-reasoned point; judging whether an abnormal point exists in a tracking process or not, marking the abnormal point as a to-be-corrected point; correcting a tracking result of a next frame based on a position of the to-be-corrected point in the Nth frame; reasoning the to-be-reasoned point to obtain a tracked result after reasoning; performing similarity comparison on the position of the tracked sample point in the N+1th frame and the position of the tracked sample point in the Nth frame, if the positions are not similar, then exchanging the tracking result of the sample point, and iterating all sample points according to the process until all sample points are rearranged to obtain a final tracking result. The application solves the problem that the existing technology cannot realize accurate matching of the sample point by simply relying on the template matching, and the problems of missing tracking points, inaccurate positions and disorder of the final sample sequence order.

This application discloses a training method for a source evaluation model, a source evaluation method, and related products. Text sample sequences and frequency sample sequences are input into the evaluation model to be trained. The model encodes the text sample sequences and frequency sample sequences to obtain text sample vectors corresponding to the text sample sequences and frequency sample vectors corresponding to the frequency sample sequences. The evaluation model then performs prediction and evaluation processing on the text sample vectors and frequency sample vectors to obtain source evaluation prediction results. Based on the difference between the source evaluation result labels and the source evaluation prediction results, the parameters of the evaluation model are adjusted until the adjusted model meets the model training cutoff condition, and training ends to obtain the source evaluation model. Thus, this application can construct a source evaluation model based on the source sample name and the title, keywords, and publication frequency of the published sample text as features, thereby achieving the evaluation of the source.

The present disclosure provides an abnormal tissue growth prediction method, device, electronic equipment and storage medium, the method comprising: obtaining an abnormal tissue growth prediction model; obtaining a historical sample sequence and a target growth duration, the historical sample sequence being arranged in chronological order by M historical samples obtained by examining abnormal tissues; performing a time sequence feature extraction operation; inputting the time sequence features of the historical sample sequence and the Mth historical sample with added noise data into a generative model, outputting post-growth image noise and post-growth abnormal tissue segmentation results; based on a preset denoising formula, using the Mth historical sample with added noise data and post-growth image noise to obtain post-growth image prediction results. In this way, the recurrent neural network for extracting time dimension information is embedded into the generative model for extracting spatial dimension information, improving the performance of the model and making the prediction results of abnormal tissue growth more accurate.

Methods and systems, including computer programs encoded on computer storage media, are provided. One example method includes: obtaining multiple pairs of sub-tensors based on first sub-tensors from a first tensor and second sub-tensors from a second tensor; determining a spatial grid size based on a size of samples, a size of the first feature dimension, a size of the second feature dimension, a size of a dimension of each first sub-tensor corresponding to the first feature dimension, and a size of a dimension of each second sub-tensor corresponding to the second feature dimension; determining whether the spatial grid size meets a threshold; and determining whether to partition the sample along the sequence length of the sample into a plurality of segments based on whether the spatial grid size meets the threshold, where each segment is assigned to one of a plurality of computing units of a processor for parallel processing.

The invention discloses an edible oil processing monitoring system, and particularly relates to the field of food processing monitoring and risk assessment, the edible oil processing monitoring system comprises a parameter acquisition module, a behavior marking module, a section division module, a difference comparison module and a risk identification module, the method comprises the following steps: continuously collecting temperature, acid value, pressure and volatile component concentration in an edible oil processing process to obtain original data streams, and sorting the original data streams into a physical time parameter sequence according to a sampling sequence; according to the method, a double-domain structure of a physical time parameter sequence and an event time behavior sequence is constructed, a response direction consistency index is extracted and compared, a response difference value and a response offset delay index are normalized, an abnormal event section is identified, and a causal chain path of the abnormal event section is traced; the problems of risk causal chain breakage and evaluation distortion of a traditional monitoring system are solved.

A training set may be generated to include a plurality of noisy sample sequences. Each noisy sample sequence in the training set may be generated by adding noise to a corresponding sample sequence from a data distribution. A protein design computation model may be trained by at least applying the protein design computation model to generate one or more output sequences, and adjusting the protein design computation model to reduce a difference between the one or more output sequences and the plurality of noisy sample sequences in the first training set. The trained protein design computation model may be applied to generate an output sequence by at least modifying an input sequence.

The application discloses a kind of based on load participation factor comprehensive energy system multivariate load prediction method, the method includes the following steps: data acquisition and pretreatment;Calculate load participation factor;Replace load data;Construct sample sequence;Build multivariate load prediction model;Multivariate load prediction.The present application extracts the hidden information between multivariate load and overall load through load participation factor, establishes neural network model to excavate the coupling characteristics between multivariate load, and improves the accuracy of multivariate load prediction.In addition, load participation factor application criteria are proposed, which improves the generalization ability of the prediction method in different application scenarios.

The application discloses a kind of based on edge computing's welding robot intelligent control method, comprising: S1, in the edge computing node collection welding process multi-source perception data, forms welding sample sequence after pre-processing;S2, calculate the intensity of sensing noise, and distribute smooth threshold parameter and sample weight;S3, welding sample is carried out real-time prediction using improved SSVR algorithm, and the predicted value of penetration, weld width and spatter amount is output;S4, set kernel function type and kernel width adaptive adjustment mechanism, calculate and update kernel width parameter;S5, calculate sliding window residual and obtain prediction uncertainty index;S6, limit the parameter update amplitude of welding current, voltage and welding speed, generate parameter adjustment instruction;S7, parameter adjustment instruction is sent to welding robot controller to execute parameter correction and trajectory fine adjustment.The application realizes the high-precision adaptive control of welding process.

The invention relates to the technical field of natural language processing, and discloses a common-situation reply generation model training method and device based on knowledge distillation. The method comprises the steps that a first sample sequence is input into a teacher model for reply generation, and a first reply sequence is obtained; inputting the emotional common sense sequence into an emotional encoder, and inputting the cognitive common sense sequence into a cognitive encoder to obtain an emotional global representation vector and a cognitive global representation vector; inputting the fusion representation vector into a decoder of the student model for reply generation to obtain a second reply sequence; performing emotion classification prediction to obtain emotion classification probability distribution; and carrying out joint training on the student model and the teacher model to obtain a common-situation reply generation model. According to the scheme, emotional changes and potential intentions of the user can be accurately captured, and personalized and diversified replies are generated in combination with a knowledge distillation technology and external common knowledge; dialogue historical information can be comprehensively and deeply analyzed, and it is ensured that generated replies are closely related to contexts, and logic coherence is achieved.

The invention discloses a middle-aged and elderly people health risk early warning and collaborative rescue scheduling method during travel, and relates to the technical field of data processing, and the method comprises the steps: 1, continuously collecting vital sign data, motion data and geographic position data of middle-aged and elderly people in a wearable terminal and a mobile terminal, generating a minute-level sample sequence of the travel health state of the middle-aged and elderly people and performing basic risk level division; 2, constructing an individual Gaussian process time sequence health risk time curve model in the server based on the middle-aged and elderly travel health state minute-level sample sequence, calculating a system condition value-at-risk index, and executing collaborative rescue scheduling decision vector search; 3, health risk early warning and cooperative rescue scheduling on the middle-aged and elderly people during travel are realized; according to the method, the accuracy of risk prediction, the sensitivity of space identification and the systematicness of a scheduling strategy can be improved at the same time, so that early warning is advanced, rescue is timely, and resource allocation is more reasonable.