32 results about "Gibbs sampling algorithm" patented technology

The invention discloses an LDA (latent dirichlet allocation) and VSM (vector space model) based similar Chinese herb literature recommendation method. The method includes: adopting an IKAnalyzer to perform word segmentation on topics and summary information of literature on the basis of a terminological dictionary for Chinese herbs, constructing a vector space, performing dimensionality reduction on the vector space, constructing a semantic dictionary, numbering all lexical items in the dictionary in sequence, performing vectorization through each document on the basis of the semantic dictionary, constructing term vectors of each document, utilizing LDA and a Gibbs sampling algorithm to perform training to obtain probability distribution of each document on themes, then computing a value of similarity between every two documents by the aid of KL divergence, computing cosine similarity of the term vectors of each document on the basis of term frequency, performing joint weighting on the two kinds of similarities prior to performing similarity sorting, and then making recommendation. By the method, the literature, similar both in content and theme, in the Chinese herb literature can be recommended to users, and recommendation results are closer to user requirements.

The invention provides a subtopic mining method. The method comprises the steps that (1) a subject value of each term of each document in a corpus is initialized; (2) based on the current subject values of all the terms of all the documents, the probability of each term in each article coming from all subtopics and the probability of each term coming from a background module are calculated, and then a subject value is redistributed for each term in each article through a Gibbs sampling algorithm based on the calculated probabilities, wherein the probability of each term coming from the background module is calculated according to term distribution vectors, subjected to statics in advance, in the background module, and the term distribution vectors in the background module are constant from beginning to end in the iteration process; and (3) if iteration stop conditions are met, LDA subtopics are obtained according to current subject value information, and if not, the step (2) is returned to. Through the method, the topic mining effect targeting a feature article set can be remarkably improved.

The invention discloses a multi-document automatic abstract generation method based on phrase subject modeling. Multiple sample documents are subjected to word segmentation to obtain phrases and frequency of occurrence of the phrases, and the documents are expressed in the form of a phrase bag; joint probability distribution of the documents is calculated on the basis of an LDA subject model, the LDA subject model is converted into a phrase subject model, then a Gibbs sampling algorithm is used for estimating implicit parameters in the phrase subject model according to Bayesian probability, and lastly probability distribution of the subject in words is obtained; the tested documents are subjected to word segmentation, the subject weight and word frequency weight of obtained sentences are calculated and obtained, the final weight of the sentences is obtained by means of weighting calculation, and abstract content is generated according to the final weight. The method is more standard and precise, the relationship between different words is taken into consideration, the subject weight of the sentences is introduced, and the generation result better conforms to the practical essay abstract writing conditions of people after the subject weight of the sentences is introduced.

The invention discloses an online topic detection method and system for a text stream. A user can quickly find interested topics in a complex text. According to the technical scheme, the method comprises the steps of building an ODT-LTF algorithm framework; extracting topics by adopting an LDA Bayesian network structure model; inferring implicit parameters of the LDA Bayesian network structure model by adopting a Gibbs sampling algorithm; building a three-order tensor of topic-topic-time through an incremental building method of the topic tensor, and fusing a time dimension in the topic tensor; decomposing the three-order topic tensor; and clustering the similar topics to obtain topics, a hierarchical structure on the topics and change of the topics in time, thereby finishing online topicdetection.

The invention provides a processing method for text clustering, a server and a system. The method comprises the steps that one theme is randomly distributed for each word in texts of a preprocessed text set from a theme set, the texts in the text set with distributed themes are distributed to multiple second servers, the initial mapping relation of each word in the texts distributed for the multiple second servers is established, clustering results of the texts in the text set are determined according to the updated theme of each word of the texts distributed for the second servers and fed back by the second servers, and the updated theme of each word is sampled and calculated by the second servers based on an improved Gibbs sampling algorithm according to the initial mapping relation of the words on the second servers. The data volume processed by the second servers and memory consumption can be effectively reduced and network bottlenecks can be avoided by determining the word mapping relation and using a matrix of a dense data structure and the improved Gibbs sampling algorithm.

The invention discloses an online social network (short for OSN) community discovery method based on a Latent Dirichlet Allocation (short for LDA) theme model. The method comprises the following steps first pre-processing data, building an LDA theme model (including an LDA-F model and an LDA-T model) based on a relationship between a user in the online social network and other friends and word information expressed by the user to solve a model probability distribution, then estimating parameters via a Gibbs sampling algorithm, and at last discovering an OSN community according to the estimated parameters. By the use of the OSN community discovery method based on the LDA Theme model, a corresponding probability model can be achieved based on user blog semantic information discovery without the use of information connection via the network topology; blog content semantic similarities are introduced to effectively describe user interest and hobby probability distribution conditions; and with the introduction of community internal topological connection tightness, communities with close internal topological connections can be discovered.

The invention discloses an improved-Gibbs-sampling-based Beidou satellite selection method comprising the following steps: calculating an azimuth and an elevation angle of a Beidou visible satellite at a monitoring point to obtain state matrixes of n Beidou satellite, and using a GDOP value as an objective function of Beidou satellite selection; and on the basis of an optimized Gibbs sampling algorithm, searching for a Beidou satellite combination enabling the objective function value to be optimized to obtain an optimal solution. On the basis of the adaptive sweeping type improvement strategy, a phenomenon of falling into the local optimization of the algorithm is avoided and convergence to the optimal Beidou satellite selection scheme is realized well. According to the invention, the optimized Gibbs sampling algorithm is applied to the Beidou satellite selection process and the inherent advantages, including low complexity, fast convergence speed and good performance of the Gibbs sampling algorithm are utilized fully, so that the quality and efficiency of the Beidou satellite selection are improved effectively. The improved-Gibbs-sampling-based Beidou satellite selection method has the broad application prospects.

The invention discloses a knuckle recognition method based on an infinite dirichlet process mixture model. The method is implemented by the following steps of 1, on the basis of a local Markov hypothesis, converting a learning problem of condition random measure into a random clustering learning problem; 2, describing probability density by using the infinite Dirichlet process mixture model, and expressing the number of clusters as a random state; 3, by utilizing a Gibbs sampling method, carrying out iterative learning on a density structure in a layered probability form; and 4, based on a collapse Gibbs sampling algorithm of the Dirichlet process mixture model (DPMM), performing model training learning by applying a sample set, and recognizing knuckles of a hand image by adopting a fixedthreshold value. According to the method, the description of the biological structure of a hand is refined; a detection result is stable and reliable; and the calculation efficiency is high.

The invention relates to a Chinese weibo sentiment analysis method based on lexical item subjective and objective directivity. The Chinese weibo sentiment analysis method comprises the following stepsof 1, obtaining a to-be-analyzed target weibo dataset; 2, conducting pre-operations like word splitting, word class tagging and stopword filtering on each weibo, and conducting combined operation onsentiment words of which the front are privatives; 3, introducing emotional transcendence knowledge and directivity transcendence knowledge on the preprocessed weibo data; 4, using a Gibbs sampling algorithm to sample the directivity, the sentiment and the subject tab of each lexical item; 5, calculating the directivity and sentiment joint distribution variable of each weibo; 6, calculating the final sentiment polarity probability distribution of each weibo, and then determining the sentiment polarity of each weibo. By means of the Chinese weibo sentiment analysis method based on the lexical item subjective and objective directivity, the conception of the subjective and objective directivity (for short, directivity) of the lexical items is put forward aiming at the weibo data, and the Gibbs algorithm is utilized to jointly model the relation of the directivity, the sentiment and the subject. The Chinese weibo sentiment analysis method is simple and practical, and the weibo sentiment analysis performance can be obviously improved.

The invention relates to a learning and reasoning method for a hybrid Bayesian network, and the method comprises the steps: S1, the discretization of a continuous variable is carried out through employing a supervised top-down discretization algorithm CACC; s2, a Gibbs Sampling algorithm is adopted to solve the parameter learning problem of an incomplete data set, and a complete hybrid Bayesian network model is obtained; and S3, a Markov chain Monte Carlo algorithm (MCMC) and a joint tree algorithm are adopted as a hybrid Bayesian network inference algorithm, and an inference algorithm more suitable for an application scene is sought. The continuous variables are discretized by adopting the CACC algorithm, so that the calculation efficiency is ensured; according to the parameter learning method of the incomplete data set adopted by the invention, the parameter learning steps are simplified, the network construction period is shortened, and the occurrence of extreme data is avoided. Thetwo hybrid Bayesian network reasoning methods adopted by the invention are good in accuracy and timeliness.

The invention provides a parameter configuration optimization method and optimization system for a cloud storage system, and the method comprises the steps: carrying out dimension reduction of parameters, selecting the parameters with the maximum impact on the system, and guaranteeing the high efficiency of parameter sampling; mining correlation between the parameters effectively through a gibbs sampling algorithm to ensure effectiveness of a data set; searching an optimal parameter configuration list by a genetic algorithm, and finally, making effective parameter configuration recommendationby a recommendation algorithm in a current system security state, so that the problem that parameter configuration comprehensive performance of an existing cloud storage platform is not obviously improved can be effectively solved; Comprehensive performance indexes are established, and the read-write and delay performance of the system is comprehensively considered.

The invention relates to a Gibbs parameter sampling method applied to a random point mode finite hybrid model. The method comprises the steps that firstly, a random point mode finite hybrid model anda random point mode likelihood function are constructed, then random point mode finite hybrid model parameter prior distribution is constructed, and posterior distribution of model parameters is obtained according to the model parameter prior distribution; and finally, estimating the number of distribution elements in mixed distribution and model parameter values by adopting a sampling algorithm combining a Gibbs sampling algorithm and a Bayesian information criterion. Compared with the traditional FMM which only describes the characteristic randomness of the data, the random point mode distribution function also describes the cardinal number randomness of the data; on the basis of RPP-FMM, a Gibbs sampling algorithm is adopted to sample sample data to obtain model parameters, and the situation that parameter estimation may fall into a local extreme point all the time, and a global extreme point cannot be obtained is avoided. According to the method, the modeling precision and the parameter estimation precision are effectively improved.

The invention relates to the technical field of text abstracts, in particular to a Gibbs limited text abstract generation method utilizing a pre-training model. According to the method, a model is used for training and generating a text abstract, and the training comprises the following steps: 1) performing word vectorization on a text source sequence, and adding a relative position code to obtain Word Embedding; 2) extracting features by using an attention mechanism and Bi-LSTM, training a model, and finely tuning the model to obtain the output of an encoder; (3) adding a relative position code to obtain a target sequence Word Embedding; (4) enabling the parameters of the decoder end to be consistent with those of the Transformer; 5) inputting the Attention matrix into a full connection layer, and then calculating to obtain probability representation of the vocabulary; and 6) fusing an LDA model into a decoder end to carry out keyword extraction, and extracting and generating an abstract in combination with a Gibbs sampling algorithm. According to the method, the text abstract can be better generated.

The invention discloses an image segmentation method of a high-order MRF model based on multi-node topology overlapping measure. The method comprises the steps of firstly, inputting a to-be-segmentednatural image; then initializing parameters; constructing a high-order MRF priori energy item based on MTOM; establishing a high-order MRF image segmentation energy model based on the multi-node topological overlapping measure according to the local region consistency WGMM likelihood energy and a region-based partial second-order Potts prior model; and optimizing the high-order MRF image segmentation energy model by adopting a Gibbs sampling algorithm, and determining an image segmentation result. According to the method, strong noise and texture mutation interference of the image can be effectively resisted, the robustness is better, and more accurate image segmentation edges are provided.

The invention discloses a competitive product level theme preference mining method. The method comprises the following steps: 1, constructing and representing a user data set; 2, modeling a competitive sub-market, a competition-related theme and a background theme; 3, modeling limited attention of a user, 4, constructing a parametric Bayesian model, and 5, carrying out parameter inference by utilizing a collapse type Gibbs sampling algorithm. When large-scale user generated contents are dealt with, the competitive sub-markets and themes corresponding to the competitive sub-markets can be effectively, quickly and accurately recognized, and enterprises can quickly recognize competitors and insight into focus topics concerned by users on competitive products.

The invention discloses a trend detection method of a parameterized hydrometeorological extreme value sequence. The method comprises the following steps: (1) establishing a trend model of the hydrometeorological extreme value sequence by adopting a generalized extreme value distribution GEV function; (2) estimating parameters of a trend model of the hydrometeorological extreme value sequence based on a Bayesian inference framework; and (3) obtaining a Bayesian log-likelihood ratio according to the parameter estimation value of the trend model, and then obtaining the trend of the hydrometeorological extreme value sequence. According to the method, the trend detection precision of the hydrometeorological extreme value sequence is effectively ensured from the parameterization angle; based on the use of a Bayesian statistical inference framework, on one hand, the estimation precision of the distribution parameters of the hydro meteorological extreme value sequence is not influenced by introducing excessive parameters under the condition of a complex trend mode, and on the other hand, the estimation process of the distribution parameters of the hydro meteorological extreme value sequence is simplified by adopting a Gibbs sampling algorithm.

The invention discloses a user personalized preference mining method based on texts and images, which comprises the following steps of: 1, constructing a user set and extracting product text description information and image information in product information purchased by users, 2, designing a parametric Bayesian model STILT (SparseTextand Image Link Topic) modeling preference content and user interest content, and 3, carrying out parameter inference by using a collapse type Gibbs sampling algorithm. According to the method, the personalized preference of the user can be effectively mined in combination with the multi-modal data of the pictures and the texts, and the user preference is focused in a certain range, so that the comprehensiveness, the accuracy and the rapidness of user personalized preference mining can be improved, accurate recommendation for the user is facilitated, and a personalized recommendation strategy is formulated.

The invention discloses a method for discovering a network public opinion theme and concerned user groups thereof. The method comprises the steps: 1, constructing a data set in a public opinion document; 2, modeling a public opinion text topic and concerning the user groups of the public opinion text topic; 3, designing a parametric Bayesian model; and 4, carrying out parameter inference by a collapse Gibbs sampling algorithm. When large-scale online social media content and social user comment behaviors are dealt with, on one hand, public opinion topics in the network can be quickly, effectively and accurately found in combination with topic analysis, public opinion detection is facilitated, and decision support is provided for public opinion guidance and control; and on the other hand, the user groups concerning each public opinion topic can be identified, and social media users can be quickly and accurately classified according to similarities and differences of concerned topics.

The invention relates to an adaptive iteration Gibbs sampling method suitable for an LDA topic model, and belongs to the technical field of computer and improved algorithm optimization. The method comprises the following steps: firstly, carrying out word segmentation and stop word removal processing on an input text data set; converting the preprocessed text data set into a bag-of-word model; inputting the word bag into the LDA topic model, and performing parameter estimation by using an adaptive iterative Gibbs sampling algorithm; and when the Gibbs sampling iteration is automatically finished, outputting potential topic features of the text data set. According to the adaptive iterative Gibbs sampling algorithm, manual setting of the number of iterations is not needed when parameter estimation of the LDA topic model is carried out, the number of iterations is greatly reduced, and the efficiency of generating topic features by the LDA topic model is improved.