15 results about "Sequence structure" patented technology

The invention discloses a method and a system for constructing a collision-zone-free frequency hopping sequence set in a '1 + M' form, and mainly solves the problems that in the prior art, frequency point collision is generated when time delay deviation is large, and new collision is introduced in a secondary mapping or expansion process. The implementation scheme of the method comprises the following steps: generating two frequency hopping sequence set sums with different collision-free regions by adopting a collision-free region construction method; combining the two sequence sets to obtain 1 + M frequency hopping sequences; segmenting 1 + M frequency hopping sequences according to a set length to obtain sequence segments of the first sequence; setting random numbers which are uniformly distributed between 0 and 1, and constructing a zeroth dynamic frequency slot set and a first dynamic frequency slot set by utilizing the sum of the frequency slot sets, the size of the total frequency slot set and the sum of frequency slot elements in the total frequency slot set; and performing frequency slot mapping by using the sum sequence segment to obtain a collision-zone-free frequency hopping sequence set consisting of 1 + M frequency hopping sequences. According to the method, effective expansion of the collision-free area can be realized while the consistency of the sequence structure is kept, and the method can be used for orthogonal frequency hopping networking when the relative time delay difference between users is relatively large.

The application relates to the technical field of key generation, in particular to a data encryption and key generation method for online payment, which comprises the following steps: extracting a touch rhythm to generate a sequence, performing round encoding to construct a bit structure, and combining amount information to generate a key. In the application, user touch behavior characteristics are converted into a rhythm sequence, a dynamic encryption basis is constructed through sequence rotation and bit mapping mode, a user behavior level parameter is introduced in the key generation process, payment behavior and transaction data are deeply bound by fusing amount section labels, the key generation no longer depends on fixed rules or external key negotiation process, independent key output on the terminal side is realized through sequence structure reorganization and template parameter matching, the individuality and unpredictability of the key are improved, the identification ability for imitated behavior and the protection effect for man-in-the-middle attack are enhanced, and the problems of static characteristics and response lag of the traditional key mechanism in actual deployment are effectively alleviated.

A knowledge injection-based fine-grained entity classification method and an interactive system are disclosed. The method comprises: inputting a sentence, a mentioned entity and knowledge graph triple information of the mentioned entity into a sentence-level encoder; constructing a feature representation of a word, a mentioned entity and triple information output by the sentence-level encoder into a graph structure and inputting the graph structure into a graph-level encoder; and inputting knowledge-enhanced sentence feature representation output by the graph-level encoder into an entity classification model for fine-grained entity classification of a target phrase in the sentence. On the basis of injecting fact knowledge in a knowledge base into a sentence in a sequential encoding manner and encoding entities based on a sequence structure, the application further constructs a graph structure using the fact knowledge and the sentence and encodes the graph structure based on GNN. Thus, the entity classification model can jointly learn entity representation learning in the sentence and in the knowledge graph structure and perform fine-grained entity classification with higher accuracy.

The invention belongs to the technical field of sequence recommendation and forgetting learning, and discloses a sequence recommendation system forgetting learning method based on structure perception, which comprises the following steps: aiming at polluted interaction data in a sequence recommendation model, evaluating influence characteristics of each polluted item through a double-index quantitative diagnosis mechanism; wherein the sequence influence range index SIS quantifies the pollution through predicting the propagation degree of a dependency relationship and the sequence structure dependency index SSD, and evaluates the structural importance of a project on maintaining the sequence time sequence coherence from global and local levels; an adaptive fine tuning strategy is executed based on the SIS and the SSD obtained through diagnosis, influence decoupling suppression pollution propagation and intensity are controlled by the SIS, adaptive interpolation is carried out between isolation and restoration strategies according to an SSD value in structure compensation, and further, the generalization ability of a model on clean data is kept by keeping loss; according to the method, targeted intervention is achieved, the sequence structure integrity is kept while effective pollution is eliminated, and recommendation performance and credibility are maintained.

The invention relates to the technical field of learning progress feedback, in particular to a dancing motion learning progress self-adaptive adjusting method and system, and the method comprises the following steps: analyzing a joint three-dimensional posture based on posture capturing equipment, obtaining a joint rotation sequence structure, calculating the rotation angle change of continuous frames, and judging a rhythm transition positioning point; and extracting a center label of the sequence structure of the analysis section, judging the distance between the center label and the transition point, adjusting the analysis section and correcting the node. According to the method, the jump position in the action is judged by comparing the direction and breadth of rotation change between continuous frames, the center mark number of the structure is extracted in the local frame segment according to the relative relation of angle change, then the offset parameter is obtained according to the interval relation between the center mark number and the jump position, and the analysis section is readjusted when the offset exceeds the range. The node recognition is established on the basis of a continuous rotating structure, so that the jump position has a dynamic reference basis, and the problem that a static node mode cannot adapt to continuous action changes is solved.

The invention relates to the technical field of key generation, in particular to a data encryption and key generation method for online payment, which comprises the following steps: extracting a touch rhythm generation sequence, executing rotation coding to construct a bit structure, and generating a key in combination with amount information. According to the method, the user touch behavior characteristics are converted into the rhythm sequence, a dynamic encryption basis is constructed through a sequence rotation and bit mapping mode, user behavior level parameters are introduced in the secret key generation process, and payment behaviors and transaction data are deeply bound through fusion of the amount section labels; key generation does not depend on a fixed rule or an external key negotiation process any more, independent key output of the end side is achieved through sequence structure recombination and template parameter matching, the individuation and unpredictability of the key are improved, the recognition capacity for counterfeit behaviors and the protection effect for man-in-the-middle attacks are enhanced, and the key generation efficiency is improved. And the problems of static characteristics and response lag of a traditional key mechanism in actual deployment are effectively relieved.

The invention discloses a scientific calculation script generation method and system. The method comprises the following steps: receiving scientific calculation task demand information described by a natural language; an encoder in a pre-trained deep learning model is adopted to encode the demand information to obtain a context vector matrix, the context vector matrix is used for representing semantic information of the demand information, and the deep learning model adopts a sequence-to-sequence structure; and decoding the context vector matrix by adopting a decoder in a pre-trained deep learning model to obtain a decoding result, and determining a scientific calculation script of the demand information based on the decoding result. The method provided by the invention at least solves the technical problem that the accuracy of the generated script is relatively low due to the fact that the script is generated based on the preset rule template in related technologies.

The invention provides a voltage quality treatment method and system based on multi-scene data, and the method comprises the steps: carrying out the timestamp matching of a multi-scene record, and extracting a continuous jump point set, so as to enable the voltage node difference, reactive power regulation action change and gear switching points to have a unified sequence structure; consistency mapping of multiple types of records on a time axis is realized; performing grade reconstruction on multiple variable quantities through a convolutional neural network response sorting mechanism, and enabling sequence differences to have a scalable processing capability; introducing an extreme value retrieval and average value sequence comparison mechanism, so that a fluctuation section can form a stable fragment set according to a difference value distribution rule; a deep neural network is adopted to calculate difference value importance, continuous large area recognition of a difference value time sequence is achieved, action conflicts are clearly expressed in a chain structure on a time axis, and a traceable algorithm-level result is provided for voltage anomaly positioning.

The invention discloses a multi-biological sequence model unified management method, system and device, a medium and a product. The method comprises the steps that in response to a biological sequence reasoning requirement, biological sequence task information is obtained, and the biological sequence task information comprises multiple pieces of sequence information; analyzing each piece of sequence information to obtain a corresponding biological sequence structure feature and a sequence type thereof; obtaining a biological sequence target model corresponding to each sequence type; grouping the multiple pieces of sequence information to obtain a plurality of subgroup sequence sets; memory requirements corresponding to each subgroup sequence set are obtained and sorted, a required memory sorting result is obtained, and batch strategies corresponding to the multiple pieces of sequence information are determined according to the current actual memory; and dynamically determining a model loading strategy according to the batch strategy, and executing a reasoning task according to the model loading strategy. According to the method, when a large amount of diversified biological sequence data is processed, appropriate models can be efficiently and intelligently selected and managed, and effective utilization of resources and efficient execution of tasks are achieved.

The invention relates to the technical field of high polymer material detection, in particular to an analysis method of a copolymer sequence structure and application thereof. According to the method, copolymer powder is subjected to high-temperature cracking under the action of a cracking catalyst, and a cracking product is subjected to gas chromatography-mass spectrometry, so that the sequence structure of the copolymer is obtained. According to the invention, by optimizing the separation conditions of gas chromatography, the determination and good separation of multiple sequence structures in the copolymer are realized. According to the method, 26 sequence structures after polyformaldehyde cracking and six sequence structures after styrene-acrylonitrile copolymer cracking can be measured and separated, the measured sequence structures are comprehensive in variety and high in separation degree, and the accuracy of sequence distribution analysis is effectively improved.

The invention discloses a simulation sequence generation method, system and equipment based on double genetic transfer matrixes. The method comprises the following steps: step 1, converting sequence data into state relation data; step 2, constructing a double transfer matrix; step 3, genetic algorithm optimization; 4, simulation data are generated based on the double matrixes; 5, outputting and applying data; according to the method, the object transfer matrix and the operation transfer matrix are constructed to realize separated modeling of the object and the operation, the genetic algorithm is combined to optimize the double matrixes generation by generation, and the object-operation legality mapping table is matched, so that the flexibility and controllability of the simulation data generation process are greatly enhanced; the generated simulation data effectively keeps the statistical characteristics and sequence structure details of the original data, the diversity of the generated data is improved, the consistency of object-operation combinational logic is guaranteed, the generated simulation data is ensured to be reasonable, large-scale real data does not need to be depended on, and the computing resource requirement is reduced.

The invention relates to a man-machine collaborative education resource recommendation method and system for large-scale online learning, and belongs to the technical field of intelligent education. The method comprises the following steps: firstly, modeling video watching interaction behavior information of a learner in a learning process, and respectively converting video interaction data of the learner into a graph structure and a sequence structure so as to describe long-term learning preference and short-term learning attention change of the learner; secondly, through a man-machine cooperation mechanism based on multiple rounds of dialogues, embedded representation of the learner and teaching resources is dynamically updated by utilizing real-time interaction feedback of the learner in the learning process, and continuous perception and adjustment of learning requirements are realized; and finally, in combination with professional background information of the learner, a recommendation strategy is adaptively optimized through a synergistic effect among different recommendation agents, so that a teaching video recommendation result which better conforms to the learning progress and personalized requirements of the learner is provided for the learner.

The application discloses a behavior sequence structured coding method based on high-order nonlinear dependence. First, a time sequence sensing embedding is performed on an original behavior sequence. Then, a network formed by stacking multiple structured coding units is used to perform hierarchical coding on the embedded sequence. Each coding unit comprises parallel processing paths, which are respectively used for processing first-order linear features and high-order nonlinear interaction features, and each order of features is weighted and combined through a learnable transformation module. Finally, the output of the network is aggregated and projected to generate a coding representation suitable for a downstream task. The application constructs a network structure comprising multiple order processing paths, so that the neural network can more effectively encode and utilize the structured high-order nonlinear dependence relationship hidden in the sequence data, and provide a more expressive feature representation for sequence modeling tasks.

The invention relates to the technical field of variable frequency regulation, in particular to a variable frequency speed regulation energy-saving control method and system based on load dynamic identification, and the method comprises the steps: constructing a multi-dimensional time sequence structure for data collected in a current operation period of a system, and identifying the dynamic trend of a current load based on the sequence; establishing a candidate frequency set, outputting a power prediction value corresponding to the candidate frequency set through a power estimation model, calculating a unit energy-saving potential after the current frequency is adjusted to the candidate frequency through an energy-saving potential algorithm, and inputting the candidate frequency corresponding to the unit energy-saving potential greater than a threshold value into a feasible frequency set; on the basis of the feasible frequency set and the corresponding unit energy-saving potential, establishing a comprehensive scoring function, and selecting an optimal target frequency through the comprehensive scoring function; and inputting the optimal target frequency and the current operation frequency into a feedback correction function, outputting an execution frequency of the frequency converter through the feedback correction function, and converting the execution frequency into a corresponding control instruction.

Contemporary gene sequencing techniques, including “Next Generation Sequencing” techniques, can include sequencing a plurality of fragments of a target polynucleotide. However, tire limitations of existing sequencing techniques, and the often repetitive or otherwise difficult-to-sequence structure of natural polynucleotides, means that it can be difficult and / or expensive to generate accurate sequences. Methods provided herein include inserting dual polynucleotide ‘barcodes,’ along with neighboring primer sequences, into a target polynucleotide prior to other sequencing processes. These inserted barcodes can improve the accuracy of sequences generated for the target by adding ‘noise’ into the target, allowing subsequent sequencing techniques (e.g., alignment, stitching, etc.) to more accurately estimate the target-plus-barcodes sequence. The primers can cause the fragments to begin at points within the target that correspond to the beginning of other sequences, facilitating the stitching of sequence ends together. The barcodes can then be removed to provide the sequence of the target polynucleotide.