22 results about "Sequence search" patented technology

The invention discloses a diffusion model reasoning acceleration method based on optimal time step sequence search and knowledge distillation, which constructs a unified search space covering a time step sequence and corresponding model architecture configuration on the premise of not performing overall fine adjustment on a pre-trained diffusion model, and searches an optimal time step sequence. And designing a knowledge distillation training method based on the optimal diffusion time step sequence obtained by searching, constructing a main distillation loss function in a discrete time step subspace limited by the optimal diffusion time step sequence, and introducing adjacent time step consistency loss to relieve discrete errors introduced by a non-uniform time step span. In the reasoning stage, a teacher model is frozen, a distilled student model is combined with an efficient sampler ACDMS based on UniPC and fused with flow matching dynamic correction, finite step sampling is carried out on an optimal time step sequence, and reasoning acceleration under generation quality maintenance is achieved.

The invention discloses a hybrid sorting method and system integrating field embedded search and keyword search, and the method comprises the steps: selecting a text embedded model of a general field, and carrying out the field fine tuning of the embedded model; performing embedding expression on the search resources by using the fine-tuned domain embedding model, and storing vector data in an index database; constructing a keyword index for the search resources, and storing the keyword index in an index database; customizing a field retrieval sequencer for hybrid embedded search and keyword search, weighting search result sequencing scores according to data use preference characteristics, and sequencing search resources based on a final scoring result; the system comprises a fine tuning module, a storage module and a sorting module. According to the method, the search engine can better understand field professional contexts and semantics and can adapt to field professional sorting requirements; the semantic correlation and the keyword correlation are considered for sorting, so that the search efficiency is high, and the sorting result is more scientific.

The invention discloses a subsequence searching method, a comparison method, a system and equipment of a gene sequence. The subsequence searching method comprises the following steps: acquiring a preset number of sequencing data of a to-be-searched gene sequence; sequentially acquiring basic group data of a search interval corresponding to each piece of sequencing data on the reference gene sequence, and searching based on an acquisition result; the search interval corresponding to each piece of sequencing data is determined based on the to-be-searched base sequence of each piece of sequencing data; repeatedly obtaining the base data of the search interval of the next search, and searching based on the obtained result to obtain a target sub-sequence; the search interval of the next search is determined according to the next to-be-searched alkali of each piece of sequencing data and the search interval of the previous search. According to the search method, the basic group data of the search interval is obtained in advance, the sequencing data is processed in batches, calculation of other sequencing data can be processed when the prefetch result is waited, and the gene search efficiency is improved.

The invention provides a memory parking loopback detection method comprising the following steps: obtaining a pre-stored global track point set, track points comprising two-dimensional coordinates and course angles; starting from the current position of the vehicle, performing inverted search on global track points, and positioning a track point closest to the vehicle as a path planning starting point; traversing the global track point set, and storing the column index appearing for the first time by taking the two-dimensional coordinates of the track points as hash keys; when encountering points with the same two-dimensional coordinates and different course angles again, judging the points as path intersection points; if the path intersection is detected, marking indexes m and n, and deleting track points between m and n to eliminate the loopback path; and generating a local path of a first preset distance based on the cut trajectory, and outputting the local path to a vehicle control system to execute tracking. According to the method, sensor data are replaced by the track point coordinates / course angles, a camera / laser radar is omitted, and the computing power demand is greatly reduced. The comparison complexity is reduced through hash table optimization, and low-computing-power chip deployment is supported.

The application discloses a user operation log data exception identification method and device, medium and product, wherein different logs in acquired user operation log data are taken as network nodes, mutual information between different nodes is calculated, a non-directional edge is determined according to the mutual information between different nodes, a non-directional minimum support tree is constructed, nodes in the non-directional minimum support tree are oriented, a preliminary minimum support tree structure is determined, node sequence groups are obtained by performing node sequence search in the minimum support tree structure, different node sequences are taken as different individuals in an initial population, the fitness of different individuals in the initial population is calculated, the initial population is updated according to the fitness of different individuals and a preset population optimization strategy, and the node sequence of an individual with the maximum fitness in the updated population is taken as a user operation exception log. The application can improve the optimization efficiency and optimization accuracy.

A siRNA inhibition rate prediction method based on antisense chain multidimensional embedding representation, in the offline stage, a deep neural network is constructed and a multidimensional feature is constructed for the antisense chain sequence of siRNA for training, in the online stage, given the target gene name, all candidate siRNA sequences are obtained, and the trained deep neural network is used to predict the real-time input of the candidate siRNA sequence and sort the silence efficiency. The present application uses one-hot encoding and pre-training RNA-FM model embedding representation features as input for the antisense chain sequence of siRNA, constructs a siRNA inhibition rate prediction model in the form of a deep neural network, improves the effectiveness and accuracy of the prediction, and fuses it with the gene expression sequence search algorithm to construct a complete process from the target gene name input to the screening of efficient siRNA sequences.

The application belongs to the technical field of natural language processing and information extraction, and specifically relates to a fine-grained entity extraction method and system for complex documents, which comprises the following steps: obtaining an OCR text box set of a document image, extracting text content and coordinate information of each text box, and fusing the text features and visual features to generate a multi-modal fusion feature matrix; calculating geometric position relationships according to the coordinate information of each text box, modeling and fusing spatial correlations by using a multi-scale kernel function to generate a spatial bias matrix; dividing all attention heads into geometric constraint heads and global free heads, calculating attention and fusing after multi-head linear projection to obtain context-enhanced features; and calculating label emission scores, combining sequence constraint values to optimize model parameters, and outputting entity extraction results through global optimal sequence search. The application does not require external pre-training data, has small parameter quantity and fast reasoning speed, and can be used for entity extraction of complex documents such as government forms and commercial bills.

The present disclosure provides a biomolecular sequence searching method and apparatus, a device, and a storage medium. The method comprises: acquiring an unknown biomolecular sequence; inputting the unknown biomolecular sequence into a sequence encoding model for sequence encoding, to obtain an unknown biomolecular sequence representation; performing similarity searching on different candidate vector sets in a biomolecular sequence vector database by means of the unknown biomolecular sequence representation, wherein candidate biomolecular sequences of candidate vectors in the candidate vector sets have a same sequence length; on the basis of the similarity searching result, searching each candidate vector set for a preset number of known biomolecular sequence representations; and screening known biomolecular sequences corresponding to the plurality of known biomolecular sequence representations for homologous biomolecular sequences corresponding to the unknown biomolecular sequence.

The invention provides a multi-target antibacterial peptide sequence optimization method and system. The method comprises the following steps: constructing a sequence search tree based on an initial peptide fragment sequence; scoring the nodes, and selecting the current node; performing point mutation operation on the current node; copying each candidate peptide fragment into a corresponding sub-population based on an improvement demand, performing local optimization to obtain new candidate peptide fragments, and evaluating each new candidate peptide fragment; scoring each node again according to the evaluation result, and reselecting the current node; and when an iteration stop condition is met, selecting a peptide fragment sequence based on the score and outputting the peptide fragment sequence. According to the application, antibacterial peptide sequence mutation is modeled into a tree structure search path, and multi-target co-evolution is introduced in a simulation stage, so that efficient coupling between global search and attribute balance is realized, joint modeling can be performed on pharmacological attributes such as antibacterial activity, host toxicity and half-life period at the same time, and efficient and refined optimization of the antibacterial peptide sequence is completed.

The invention discloses an interrupt drive program atomicity violation detection method based on a partial order guide event graph, which comprises the following steps: acquiring a symbolized memory event in a program to be detected, and extracting a partial order of the symbolized memory event; based on the partial sequence, identifying a potential atomic violation event group from the symbolized memory events to obtain a key rf sequence; generating a skeleton of the event graph based on the partial sequence, and expanding edges of the event graph to obtain a stable event graph; and searching the stable event graph according to the key rf sequence to obtain an actually occurring atomic violation event. According to the method provided by the invention, the event graph of the SME is generated through the partial sequence, and the event graph is searched based on the rf sequence of the potential atomicity violation event group to detect the actually occurring atomicity violation event; the partial order relationship between events is visually represented, so that the interpretability of the detection process can be improved; moreover, the implementation cost is low, the adaptability is good, and the detection precision is high.

The application discloses a search method and system for a Costas sequence based on multithreading, and the method comprises the following steps: inputting the order of the Costas sequence to be searched; establishing a framework data structure of a bidirectional linked list, and generating a permutation matrix corresponding to the order through recursive iteration; generating a task for judging a check matrix corresponding to the permutation matrix, and submitting the task; submitting the task received by S2 and S3, creating a thread pool by using a thread pool technology, and performing multithread operation by combining a task stealing technology, so as to accelerate the judgment on the check matrix; and collecting the tasks, finally obtaining the number of the Costas sequences and the search time, and comparing with a traditional multithread search. The search method and system for the Costas sequence based on the multithreading realize efficient utilization of threads, improve the search speed, and provide a wider application prospect.

The invention discloses an optimal odd complementary sequence pair searching method, a framework, a storage medium and equipment, and belongs to the crossing field of signal processing and artificial intelligence, and the method comprises the following steps: initializing a sequence prefix, and carrying out searching through the steps of MCTS selection, expansion, evaluation and return; according to the evaluation step, CNN is adopted to directly carry out value estimation on a current sequence prefix, and high calculation overhead caused by random completion in a traditional MCTS is avoided; state, strategy and return information generated in the search process are subjected to independent management and fixed-length processing through a playback pool, and the stability and generalization ability of CNN training under input of sequences of different lengths are guaranteed; cNN parameters are updated in a closed-loop mode, and a search strategy is continuously optimized. In the fields of radar signal design, wireless communication channel estimation and synchronization and the like needing high-performance sequences, the optimal odd complementary sequence pair meeting the Cramer-Rao lower bound can be efficiently and stably searched, and the sequence searching efficiency and the engineering practicability are remarkably improved.

This invention discloses a dynamic phase sequence spherical decoding multi-step predictive control method for common-mode voltage suppression. It proposes incorporating a hard constraint mechanism for common-mode voltage during the search process, effectively suppressing common-mode voltage by pruning high common-mode voltage branch nodes, thus overcoming the problem of excessively high common-mode voltage in traditional spherical decoding predictive control. The method's computational reduction effect becomes more significant with increasing prediction step size. A dynamic phase sequence search strategy is proposed, periodically rotating the search phase sequence to achieve a uniform distribution of inverter switching frequency and reduce the risk of local overheating of the converter. This invention also proposes setting a hard constraint on the upper bound of the number of search nodes for spherical decoding predictive control, which can limit the algorithm's search time as needed and avoid a surge in the number of search nodes under operating condition disturbances. Simulation results based on a three-phase five-level active neutral-point clamped inverter verify that this method can achieve uniform switching frequency distribution, common-mode voltage suppression, capacitor voltage balance, and significantly reduce computational load.

The application discloses an application method of a sequence search tool CircBLAST considering gene sequence evolution rearrangement, and belongs to the technical field of bioinformatics. The method flow comprises the following steps: firstly, all protein sequences are cut according to the length of a required word_size, a data set is constructed in combination with sequence annotation data, and is written into a database; then, a request sequence is prepared, and is cut into small fragments of word_size; further, a search matching, construction of a circular sequence and sequence alignment are carried out to complete a retrieval process; finally, an alignment result containing matching fragments, a similarity score and the like information is generated, is used for presenting to a user for viewing and judging the reliability of matching. The application considers the evolution rearrangement of gene sequences, significantly improves the accuracy of sequence alignment, and can find more reordered sequences.

The application provides a menu data acquisition method and device, relates to the technical field of data processing, and can be applied to the financial field or other technical fields.The method comprises the following steps: in response to an execution action of a target menu sequence searched by a user through a menu keyword, obtaining a candidate menu list corresponding to the menu keyword; calculating the similarity between each character data in the candidate menu list and the menu keyword to obtain a first menu sequence similar in semantics to the menu keyword; and generating the target menu sequence according to a second menu sequence containing historical click rate data corresponding to the user and the first menu sequence obtained in advance.The device executes the above method.The menu data acquisition method and device provided in the embodiment of the application can quickly and accurately find the menu data required by the user.

The application discloses a kind of optimal odd complementary sequence pair search method, framework, storage medium and equipment, belong to the cross field of signal processing and artificial intelligence, which comprises: initialization sequence prefix, search is carried out by the selection, expansion, evaluation, back transmission steps of MCTS;Wherein, evaluation step uses CNN to directly estimate the value of current sequence prefix, avoid the high computing overhead caused by random completion in traditional MCTS;The state, strategy and return information generated in the search process are independently managed and fixed length processing by replay pool, guarantee the stability and generalization ability of CNN training under different length sequence input;And update CNN parameter in closed loop mode, continuously optimize search strategy.The application can efficiently and stably search out optimal odd complementary sequence pair satisfying Krammer lower bound in the field of radar signal design, wireless communication channel estimation and synchronization, etc., significantly improve the efficiency and engineering practicability of sequence search.

A wind-solar carbon flow correlation analysis method and system based on PLHS sampling and sample optimization, comprising: constructing a wind-solar output probability distribution model; constructing a system carbon flow rate comprehensive analysis model under the coupling uncertainty of wind-solar output; using a hierarchical Latin hypercube sampling method to sample wind speed and light intensity to obtain a sample sequence; an adaptive sequence search algorithm is proposed, a composite index based on spatial dispersion is used as an optimization target to optimize the sample sequence, and the sample sequence is input into the wind-solar output probability distribution model to obtain a wind-solar output sample set; traversing the sample points in the wind-solar output sample set, the system carbon flow rate and its probability distribution characteristics are calculated; for each sample point in the wind-solar output sample set, the change amount of the system node, branch and network loss carbon flow rate caused by the wind-solar output injection is calculated respectively, and the weighted average of the change amount is obtained to obtain the average influence factor of the wind-solar output on the node, branch and network loss carbon flow rate. The present application realizes reliable evaluation of system carbon emission.

The invention relates to the technical field of aerospace assembly, in particular to an interactive aircraft strength test assembly body assembly sequence rapid planning method which comprises assembly body assembly state extraction, an ant-grey wolf heuristic disassembly sequence searching method and a single-step optimal principle disassembly sequence calculation method. Wherein the assembly state extraction is used for extracting required data from an assembly model and storing the data according to a disassembly sequence evaluation requirement; an ant-grey wolf heuristic method is used for generating a feasible disassembly sequence according with an evaluation criterion according to the sequence evaluation criterion and by utilizing the extracted model assembly information; according to the single-step optimal principle method, a new assembly sequence is quickly generated according to a given evaluation criterion, model assembly information and additional sequence constraint conditions. According to the method, the advantages of the two sequence generation methods are comprehensively utilized, the intention of a designer is fused into an algorithm result in an interactive mode, and the high-quality assembly sequence of the assembly is rapidly and flexibly generated through iteration.

The application discloses a diffusion model reasoning acceleration method based on optimal time step sequence search and knowledge distillation, which constructs a unified search space covering time step sequences and corresponding model architecture configurations without overall fine-tuning of a pre-trained diffusion model, and searches for optimal time step sequences. Based on the optimal diffusion time step sequences obtained through the search, a knowledge distillation training method is designed, a main distillation loss function is constructed within a discrete time step subspace defined by the optimal diffusion time step sequences, and an adjacent time step consistency loss is introduced to relieve the discrete error introduced by a non-uniform time step span. In the reasoning stage, the teacher model is frozen, and the distilled student model is combined with an efficient sampler ACDMS based on UniPC and fusing flow matching dynamic correction to perform limited step sampling on the optimal time step sequence, so that reasoning acceleration is realized under the condition of keeping generation quality.

The invention discloses a vulnerability utilization path speculation and analysis method, system and device based on a knowledge graph, and a medium. The method comprises the following steps: encoding a security entity in a network into a heterogeneous graph structure, and constructing a heterogeneous security entity knowledge graph; constructing an inference rule set of an attack path based on the knowledge graph, wherein each rule is composed of a precondition set, an operation semantic tag and a state change set; performing action sequence search based on inference rules in the knowledge graph to generate all paths meeting attack feasibility, and performing multi-dimensional risk quantitative evaluation on each path to obtain a high-risk attack path set; and executing incremental updating and path re-verification of the graph structure, identifying an affected path, re-verifying the precondition and the risk score of the affected path, and constructing a safety analysis closed loop of continuous evaluation and self-adaptive optimization. According to the method, mining, visualization and dynamic updating can be carried out on the high-risk path, and the method has good real-time performance, expandability and intelligent capability.

A method and apparatus for predicting and analyzing RNA-binding residues are disclosed. The method constructs a GDRBind model using PyTorch and the DGL framework; inputs a protein sequence of a predetermined length into the GDRBind model and obtains the protein structure data of the protein sequence; generates a multiple sequence alignment (MSA) file using the protein sequence search software HHblits; collects several RNA-binding protein sequences from the UniProtKB protein database, performs clustering processing to obtain a pre-training dataset, and uses it to train a general protein language model to obtain an ESM-RBP representation model; obtains a first embedding matrix and a second embedding matrix through processing, and concatenates them to obtain a residue node feature representation matrix; calculates the edge features of all residue pairs in the protein sequence to obtain an edge set; and predicts the RNA-binding residues of the protein sequence using an equivariant graph neural network (EGNN) prediction model, outputting the predicted results. This invention can solve problems such as insufficient domain feature mining, low segmentation accuracy, and lack of interpretability.

The application provides a multi-target antibacterial peptide sequence optimization method and system, and the method comprises the following steps: constructing a sequence search tree based on an initial peptide segment sequence; scoring nodes and selecting a current node; performing a point mutation operation on the current node; copying each candidate peptide segment to a corresponding sub-population based on improvement requirements, performing local optimization to obtain new candidate peptide segments, and evaluating each new candidate peptide segment; re-scoring each node and re-selecting the current node according to the evaluation results; and when an iteration stop condition is met, selecting a peptide segment sequence based on the score and outputting the peptide segment sequence. The application models antibacterial peptide sequence mutation as a tree structure search path, introduces multi-target cooperative evolution in the simulation stage, realizes efficient coupling between global search and attribute balance, can simultaneously model pharmacological attributes such as antibacterial activity, host toxicity and half-life, and completes efficient and fine optimization of the antibacterial peptide sequence.