54 results about "Domain prediction" patented technology

The invention discloses a milling stability domain prediction method under multi-modal coupling to solve a technical problem that the efficiency of an existing milling stability domain prediction method is low. The technical scheme is that firstly the transmission function of a process system is determined through a modal test experiment, then modal parameters are extracted from the transmission function and two orthogonal direction modal parameters are paired and combined, a milling test is carried out to calibrate a milling force coefficient, based on the milling force coefficient and each scale of modal parameters, an improved semi discrete method is employed to obtain a stability lobe diagram under each scale of modals, and finally lobe diagrams under obtained from each scale of modals are drawn in a same coordinate system to obtain a stable domain under multi modal coupling. Through a test, when a cutting cycle is divided into 40 sections, 80 sections and 120 sections, compared with a background technology method, the method can save time for 4949.6 seconds, 74200.4 seconds and 344699.5 seconds, and the efficiency is increased by 85.9%, 92% and 88.5%.

The invention belongs to the technical field of computer network data prediction, and discloses an academic field data correlation prediction method based on deep learning, and a computer. The methodcomprises steps of collecting public general data, and paper and patent data in the academic field; training an academic field word vector on the academic corpus by utilizing a word vector technologyof deep learning; for a given field, predicting other fields related to semantics according to the word vectors, and achieving prediction of related academic fields. The system comprises a data collection module used for collecting public data; the word vector training module is used for training academic field word vectors on the academic corpus by utilizing a word vector technology of deep learning; and the academic domain prediction module is used for predicting other semantically related domains according to the word vectors for a given domain to realize prediction of the related academicdomains. According to the method, the academic field word vector based on deep learning is constructed, and field-related rapid and accurate prediction is realized by means of the word vector.

The invention discloses a Gini exponent-based domain adaptive semantic segmentation method, which comprises the following steps of: measuring uncertainty of output prediction by utilizing a Gini exponent, performing uncertainty measurement and constraint on output prediction of a target domain on an output layer, reducing the difference between a source domain and the target domain in category distribution, and performing inter-domain adaptation; dividing a target domain sample set into two subsets according to an uncertainty measurement result of the Gini index for target domain prediction, training an intra-domain adaptive segmentation network for a sample corresponding to intra-domain high-confidence prediction by using a pseudo tag as weak supervision information, and calculating a Gini index graph for output prediction of the two subsets of the intra-domain adaptive segmentation network; a Gini exponent graph calculated by a low-confidence sample is constrained, a discriminator Dtis used for discriminating which subset the Gini index graph belongs to, the difference in a target domain is reduced based on an adversarial thought, and the semantic annotation precision is improved. Compared with the prior art, the semantic annotation accuracy of the target domain is remarkably improved.

The invention discloses a commutation failure prediction method based on direct current finite time domain prediction. According to the method, after a system operates stably, a direct-current voltagedifference [delta] U and a derivative d [delta] U/dt thereof on two sides of rectification and inversion of a direct-current line are detected in real time; when the change is larger than a preset threshold value, a direct-current variable quantity critical value is calculated, and meanwhile, an iterative function is constructed to calculate a direct-current prediction value xk through a Steffensen iterative algorithm. And when the direct current prediction value xk is greater than the critical direct current variation quantity [delta] Id *, a commutation failure prediction signal is sent out. The prediction method can achieve the successful prediction of the first commutation failure, does not need to additionally increase equipment or change a topological structure of a converter, and is easier to achieve.

The invention provides a high-speed machining stability domain prediction method based on cutter tooth cutting time course fine integration. The high-speed machining stability domain prediction methodcomprises: firstly, establishing a high-speed milling dynamics model considering regeneration flutter; secondly, dispersing the rotation period of the main shaft, and simplifying the Duhamel integralin the time delay differential equation into an integral of a state item and a time delay item at each dispersed small interval; secondly, using a piecewise cubic Hermite interpolation polynomial tofit a time lag item, and meanwhile, using a plurality of known time node responses and derivatives thereof to approach a required item, so that the fitting error of a calculation method is reduced, and the precision of the prediction method is improved. In addition, in the process of solving the stable lobe graph, a fine integration algorithm is applied, matrix inversion operation in the calculation process is avoided, the time consumed by the calculation method is greatly shortened, and the calculation efficiency is improved.

The invention discloses a double-crane system variable amplitude angle response modeling algorithm and a random response domain prediction method. The variable amplitude angle response modeling algorithm includes the steps: 1 building a system geometric model and giving coordinates of points; 2 building a system constraint equation; 3 building a variable amplitude angle response model according to the system constraint equation. The random response domain prediction method includes the steps: 1 describing uncertainty of a load according to a random parameter model; 2 building a random variable amplitude angle response equivalent equation based on the variable amplitude angle response model and the random parameter model; 3 proposing a perturbation random composite function method according to composite function characteristics and a random perturbation method, and solving random variable amplitude angle response expression; 4 further solving variable amplitude angle response expectation and variance. The prediction problem of a lower variable amplitude angle response domain with random parameters can be solved, the algorithm and the method have the advantages of high speed and precision, and system operation reliability is ensured.

The invention belongs to the object tracking control field, and discloses a two-dimension track tracking controller of the underwater robot. The two-dimension track tracking controller is realized through steps of utilizing a least square method to perform modeling on a movement track under a motion coordinate system, designing a self-adapting wide domain prediction control two-dimension track accuracy tracking controller of the underwater self-adapting robot on the basis of that, and realizing the two-dimension track tracking control of the underwater robot. The invention is applicable to the object tracking and identification.

The invention provides a milling stability domain prediction method using a Newton and Hermite interpolation method, the time delay period is equally divided into a plurality of discrete cells; and then, a third-order Newton interpolation polynomial and a segmented third-order Hermite interpolation polynomial are adopted for respectively interpolating a state item and a time lag item in a time lagdifferential equation on each cell, and required items are approximated by adopting the same function value on the node and the same derivative value on the node, so that the fitting error of a calculation method is reduced. According to the method, the calculation precision and the calculation efficiency can be effectively improved, so that the milling stability domain can be predicted more accurately and efficiently, and a theoretical basis is provided for selecting reasonable cutting parameters for cutting technicians to improve the surface quality of parts.

The invention relates to a deep learning multi-step irradiance prediction method based on a coder-decoder, and belongs to the technical field of photovoltaic power generation. The prediction method comprises the following steps: S1, training data acquisition: acquiring historical irradiance data of a target area and corresponding meteorological data, and making a supervision data set; s2, data preprocessing, including meteorological information feature coding and data normalization; s3, a coder and decoder model is trained, a coder model is composed of a TCN and LSTM cascade structure, and a decoder is composed of an LSTM and MLP cascade structure; the method comprises the following steps: training a coder-decoder model by using read irradiance of a current time period t0-tN as supervision information and historical irradiance and meteorological information before a t0 moment as input data; and S4, prediction: inputting historical data into the coder-decoder model obtained by training in the step S3, and predicting the solar irradiance of multiple steps in the future. According to the method, the historical information of the irradiance sequence can be fully utilized, and experiments show that the method can effectively improve the precision of multi-step irradiance prediction.

The invention provides an algorithm for collaboratively filtering agricultural materials on the basis of time and region granularity. The algorithm includes steps of building time and region granularity matrixes by the aid of existing agricultural statistical law research results; acquiring user behavior data; partitioning the acquired user behavior data according to the time and region granularity matrixes; building user commodity score matrixes of the time and region granularity where users are located; computing user similarity according to the user commodity score matrixes; finding out the most similar N users and recommending the agricultural materials to the current users on the basis of interest of the most similar N users. Compared with the traditional algorithm for collaboratively filtering agricultural materials on the basis of users, the algorithm has the advantage that the shortcomings of low prediction accuracy of the existing algorithm for collaboratively filtering the agricultural materials and excessively huge user commodity score matrixes when similarity is computed at present can be overcome by the aid of the algorithm.

The invention discloses an inter-frame image parallel coding method based on time-space domain prediction, which relates to the technical field of image processing, and comprises the following steps: setting a space domain reliability threshold and a time domain reliability threshold according to a data compression requirement, and obtaining a motion vector candidate list of a current frame image coding block group; acquiring a space domain reliability parameter and a time domain reliability parameter according to the motion direction of the current frame image; selecting a more appropriate motion vector in the current coding block as a shared motion vector of all coding blocks in the parallel region according to the proportion of the spatial domain reliability parameter and the time domain reliability parameter and the size relationship between the spatial domain reliability parameter and the time domain reliability parameter and the spatial domain reliability threshold and the time domain reliability threshold; and according to the shared motion vector set of each coding block, parallel coding of the coding blocks in the coding block group is carried out. According to the method, the texture complexity of the image needing to be processed during parallel coding is reduced through submergence of the depth of the coding block, so that the coding block realizes parallel coding while the bit rate loss is low.

The invention discloses a prediction method for a forwarding cascade scale. According to the method, a GAT (graph attention network), a dynamic route and an LSTM (long short-term memory recurrent neural network) are combined into a new model for predicting the information cascade scale. The method comprises the following steps: preprocessing data; the graph attention network extracts node features; cascading snapshots are divided; dynamically routing and aggregating node information; the LSTM extracts time information; carrying out final prediction by using an MLP (Multilayer Perceptron); and testing the model. The method can be applied to the field of popularity prediction. The invention provides a new feature extraction method for the field, solves the problems of low prediction efficiency, low information utilization rate after multiple models are combined and the like in the field, and meets the requirements of rapid prediction and capture of hot events.

The invention provides a sample transfer learning method and device, and the method combines the prediction results of a prediction model of a source field and a prediction model of a target field ona sample in a sample transfer process, and gives full consideration to the impact on the prediction precision of the prediction model of the target field from the sample of the source field. Meanwhile, in the sample migration iteration process, the prediction models of the source domain and the target domain are continuously optimized based on the intermediate state sample set of the source domainand the target domain obtained by migration; and the performance of the prediction model of each iteration is not inferior to the performance of the prediction model in the last iteration, so that the finally migrated sample set of the target field is more beneficial to improving the performance of the prediction task of the target field.

The invention relates to artificial intelligence, is applied to smart cities, and uses natural language processing to improve search precision. The invention specifically discloses a content search method and device, a domain prediction model training method and device and a storage medium, and the search method comprises the steps: obtaining a search text, and carrying out the word segmentation of the search text to extract a plurality of keywords; carrying out embedding processing on the keywords to obtain keyword vectors of the keywords; based on the trained part-of-speech tagging model, determining the part-of-speech of each keyword according to the keyword vector; based on the trained domain prediction model, determining whether the keyword is a domain word according to the keyword vector; determining a weight value of each keyword according to the part-of-speech of the keyword and whether the keyword is a domain word, the weight value of the keyword which is the domain word beinggreater than the weight value of the keyword which is not the domain word; and based on the search engine, outputting a search result according to the keyword and the weight value thereof. The invention further relates to the field of the block chain, and the trained field prediction model can be stored in the block chain node.

The invention discloses a flexible cable driven waist rehabilitation robot state interval response domain prediction method. The method comprises the following specific steps: 1, establishing an anglestate response equation in a training process according to a flexible cable driven waist rehabilitation robot geometric model; 2, converting the angle state response equation into an angle state response equivalent equation; 3, establishing an angle state response interval equivalent equation; step 4, approximately expanding an interval vector and an interval matrix at a mean value of the introduced interval vector; step 5, obtaining an angle state response midpoint value and a change interval radius according to a perturbation theory, and step 6, obtaining upper and lower bound values of theangle state response interval vector; and 7, combining the step 5 and the step 6 to obtain an angle and stress state interval response domain. On the basis of solving the problem of redundant mechanism solving, the problem of angle and stress state response domain analysis of the flexible cable driven waist rehabilitation robot system under the condition of containing uncertain interval parameters can be solved, the calculation efficiency is effectively improved, the calculation precision is guaranteed; meanwhile, the calculation process is greatly simplified, and the calculation time is greatly shortened.

The invention discloses a 3D video intelligent multi-domain joint prediction coding method and device. The method comprises the following steps: (1) obtaining multi-domain reference information: taking reconstructed pixel regions on the left side, above and above the left side of a current coding block in a step length range as spatial domain reference information, taking an inter-frame predictionblock of the time domain correlation of adjacent frames as time domain reference information, and taking a viewpoint synthesis prediction block obtained by a viewpoint synthesis prediction technologyas inter-viewpoint reference information; (2) constructing a time-space prediction network, and obtaining a time-space domain prediction result by taking time-space domain reference information as input; and (3) constructing a multi-domain joint prediction network according to the time-space domain prediction result and the viewpoint synthesis prediction block to obtain a final multi-domain prediction result. The device comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, and the processor implements the method steps when executingthe program.

The embodiment of the invention discloses a prediction method, device and equipment for the technical hotspot field and a storage medium, and the method comprises the steps: obtaining a historical technical literature set which comprises at least two historical technical literatures; determining the technical field corresponding to each historical technical literature; for each technical field, calculating each technical evaluation index value in a technical evaluation index set according to the historical technical literature set, and determining the technical evaluation index set by a technical evaluation index set determination model generated based on machine learning training; and predicting the technical hotspot field according to the technical evaluation index values corresponding to the technical fields. According to the embodiment of the invention, comprehensive and accurate information can be obtained through machine learning, so that the accuracy of the technical evaluationindex set determined by the technical evaluation index set determination model generated based on machine learning training is relatively high. Based on this, the technical hotspot field is predictedaccording to the technical evaluation index set, so that the prediction accuracy of the technical hotspot field is improved.

The invention discloses a system for predicting whether an embryo can be capsulated or not based on a time-delay camera system and a deep learning algorithm. The system comprises a single-cell and multi-cell identification module, a pronucleus identification module, a pronucleus disappearance identification module, a five-classification cell identification module, a time domain prediction module, a space domain prediction module and a time-space domain combination prediction module. According to the system for predicting whether the embryo can be capsulated or not based on the time-delay camera system and the deep learning algorithm, time sequence and form information of a video are separately identified by using a time domain model and a space domain model, and then the time sequence and form information are integrated, so that the prediction accuracy is remarkably improved.

The invention discloses a Wasserstein distance-based similar adversarial network characterization model. Marginal probability distribution of source domain subjects and target domain subjects is reduced to the greatest extent by a method of reducing the Wasserstein distance, conditional probability distribution is reduced by a correlation enhancement method, namely, the internal relation of categories is enhanced, the scheme includes the following steps: performing sampling, filtering noise, performing mapping, setting a Wasserstein distance of a domain obfuscator, setting a gradient penalty of the domain obfuscator, adopting an association enhanced classifier, solving the similarity of feature representation from a source domain to a target domain, solving the similarity of feature representation from the target domain to the source domain, obtaining the round-trip probability of features in the source domain and the target domain, calculating the label probability of the source domain, calculating the loss of Lzw and Psts through cross entropy loss, setting the access probability, setting the target domain label probability, calculating the loss of Lop and Pv through cross entropy loss, setting the classifier loss, setting the source domain prediction classification loss, setting the number N of iterations, and when the number of times of training reaches the set number of iterations, stopping operation.

The embodiment of the invention provides an image segmentation method and device, equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: extracting target image features of a target domain to-be-processed image through a trained target domain image segmentation network, obtaining a target domain prediction image based on the target image features, wherein the target domain prediction image comprises a target object with a target label, and the trained target domain image segmentation network is obtained by performing joint iterative training on a to-be-trained target domain image segmentation network and a to-be-trained source domain image segmentation network based on a domain transfer learning mode, and the source domain sample image comprises a source domain sample object with a source label. Through the domain transfer learning mode, the target domain image segmentation network used for segmenting the target domain image is obtained under the condition that each pixel in the target domain sample image does not need to be subjected to refined labeling, so that the manpower and the cost are reduced, and the image segmentation efficiency and the image segmentation accuracy are improved.

The invention relates to a near-infrared spectrum model migration method based on a deep Bi-LSTM network, which belongs to the technical field of near-infrared model transfer. The method comprises the steps of obtaining spectral data of a source domain and a target domain, performing data enhancement on the source domain spectral data, preprocessing the spectral data of the source domain and the target domain, dividing spectral data of a source domain and a target domain, designing a Bi-LSTM network structure, training a Bi-LSTM network structure by using the source domain spectral data, extracting all Bi-LSTM layers, and adding a full connection layer to form a neural network, training a full connection layer by using near-infrared spectral data of a target domain correction set and a verification set, and updating weights and deviations among layers of the neural network, and testing the migration model by using the near-infrared spectral data of the target domain prediction set, and evaluating the migration effect and the anti-noise capability of the model. According to the method, migration from the target domain quantitative model to the source domain quantitative model is achieved, a large amount of time for reconstructing the model is saved, and high-precision prediction is kept.

The invention discloses a joint recommendation system based on a joint learning framework, which belongs to the technical field of joint recommendation systems and comprises a target domain input end, a main LSTM (Long Short Term Memory) layer, a main hidden layer, a merging layer, a main activation function layer, a source domain input end, an auxiliary LSTM layer, an auxiliary hidden layer, a joint hidden layer, an auxiliary activation function layer, an auxiliary task output end and a main task output end. The target domain input end receives related data predicted by a target domain, generates a corresponding main task according to the related data and sends the main task to the main LSTM layer, the main LSTM layer receives the main task and enables the main task to enter an LSTM network to participate in judgment of a word segmentation NER model, and meanwhile, the main LSTM layer generates a main hidden layer on the basis of the target domain. Through a joint learning method, the source field sample assists the target field to perform word segmentation, so that the sample labeling time is shortened, and the word segmentation performance can be improved at the same time.