50 results about "Conditional independence" patented technology

Improved techniques are disclosed for accessing content in file systems, allowing file system clients to realize advantages of file system referrals even though a file access protocol used by the client is not specifically adapted for referral objects. (For example, the client may have a legacy file system protocol or a proprietary file system protocol which does not support referrals.) These advantages include a uniform name space view of content in a network file system, and an ability to locate content in a (nearly) seamless and transparent manner, even though the content may be dynamically moved from one location to another or replicated in different locations. A file system server returns a symbolic link in place of a referral, and an automated file mounting process on the client is leveraged to access the content using the link. Built-in crash recovery techniques of the file system client are leveraged to access moved content.

According to an aspect of the invention, a method for training a classifier for classifying candidate regions in computer aided diagnosis of digital medical images includes providing a training set of annotated images, each image including one or more candidate regions that have been identified as suspicious, deriving a set of descriptive feature vectors, where each candidate region is associated with a feature vector. A subset of the features are conditionally dependent, and the remaining features are conditionally independent. The conditionally independent features are used to train a naïve Bayes classifier that classifies the candidate regions as lesion or non-lesion. A joint probability distribution that models the conditionally dependent features, and a prior-odds probability ratio of a candidate region being associated with a lesion are determined from the training images. A new classifier is formed from the naïve Bayes classifier, the joint probability distribution, and the prior-odds probability ratio.

A processor having improved branch prediction accuracy includes at least one execution unit that executes sequential instructions and a plurality of branch prediction circuits including a lock acquisition branch prediction circuit that predicts a speculative execution path for a conditional branch instruction. The processor further includes a selector that selects the speculative execution path predicted by the lock acquisition branch prediction circuit in response to an indication that the conditional branch instruction is dependent upon lock acquisition. In a preferred embodiment, the indication that the conditional branch instruction is dependent upon lock acquisition is encoded within the conditional branch instruction.

A method (and system) of predicting an unobserved target variable includes building a graphical predictive model from domain knowledge, which takes advantage of conditional independence to facilitate inference about the unobserved target variable, given observations of other variables in the graphical predictive model from a plurality of information sources.

A computer readable medium, system, apparatus and method are disclosed for generating and regenerating query results in reports whereby conditions for filtering a query against multilingual databases are preserved independent of language and / or locale. According to one embodiment of the present invention, a computer readable medium includes executable instructions to specify a language-dependent value for filtering query results during a query. Other executable instructions are included to associate a first locale to the language-dependent value, determine a key based on the language-dependent value and on the first locale, and generate a transformed query, which can include a locale variable configured to indicate a second locale for regenerating the query results to form regenerated query results for the second locale using another language-dependent value.

The present invention provides a method to construct stable, high-order explicit discretization for the wave equation based on the discretization of the evolution formula. The present invention provides independent computation of discretization in one, two, or more spaces for bulk propagation, near boundaries propagation, and discretization of a projection operator to enforce boundary conditions. More specifically, the method includes an act of discretizing propagation operators LΔt by using an identity that is derived using central differencing in time. The method also includes an act of providing a high-order discretization of a boundary projection operator that enforces boundary conditions independent of the discretization of the propagation operators LΔt. Additionally, the method includes an act of alternating an application of a discretization of an evolution formula having a spatial filtering operator LΔt with a boundary projection operation for stepping forward in time to determine wave propagation in the media.

The invention has the object of obtaining, at a high rate of success, graphs indicating the relationships between variables indicating the states of observed items which are the subjects of data mining, and improving the reliability of the outputted graphs. A method for generating a graph showing the relationships between variables comprises a step S2 of establishing a number of graphs to be generated, a step S5 of randomly establishing an order of variables X forming the set of all variables V, a step S6 of performing a process of reconstructing a graph showing the relationships between variables, and a step S10 of outputting a comprehensive graph including all edges existing in any of the graphs generated with each graph generation. In the graph reconstruction process, an inverse matrix of the correlation coefficient matrix is calculated, and the operation of determining the conditional independence relating to two variables which are the subject of the conditional independence determination is skipped if any of the diagonal elements relating to the two variables is greater than a predetermined threshold value.

A system and method for detecting one or more anomalies in a plurality of observations is provided. In one illustrative embodiment, the observations are real-time network observations collected from a stream of network traffic. The method includes performing a discrete decomposition of the observations, and introducing derived variables to increase storage and query efficiencies. A mathematical model, such as a conditional independence model, is then generated from the formatted data. The formatted data is also used to construct frequency tables which maintain an accurate count of specific variable occurrence as indicated by the model generation process. The formatted data is then applied to the mathematical model to generate scored data. The scored data is then analyzed to detect anomalies.

The invention discloses a naive Bayes classification model improvement method based on attribute weighting, and relates to the field of data processing and classification. The method comprises the following steps of S1, preprocessing the data; S2, calculating a grouping Spearman coefficient, removing the redundant attributes, and updating the data set; S3, solving the prior probability and the class condition probability of each class; S4, calculating a weighting coefficient of each attribute of the updated training set; and S5, performing classification according to the weighted improved model, and performing statistics on a classification result. According to the method, the conditional independence assumption of the naive Bayes classification model is effectively weakened through an attribute weighting mode, the redundant attributes are removed through the Spearman coefficient, and the accuracy and efficiency of the naive Bayes model are obviously improved through the improved model.

The invention provides a method for the synergetic fusion of a distributed sensor network and the positional correction of a sensor. According to the method, a to-be-solved joint posterior distribution is approximated into the product of multiple distributions mutually independent through conditions, so that to-be-solved joint posterior distribution can be solved through an iteration VB algorithm.Meanwhile, the iteration process of a consistent algorithm is introduced into a variation iteration process, and the estimation of each node to a target state is restrained into an overall consistentresult through local communication between each sensor node and an adjacent neighbor node. A simulation result shows that the estimation precision of each sensor node to the overall target state andthe local position in the distributed sensor network can be effectively improved. The method can be applied to the fields of distributed sensor network radars, infrared target tracking, mobile robot positioning and the like.

The invention discloses an evaluation index equilibrium state analysis method based on Bayesian causal network, comprising the following steps: establishing a system evaluation index system, and determining exogenous factors affecting the system; obtaining a corresponding endogenous variable set from the evaluation index, The exogenous input variable set is obtained from the exogenous influencing factors of the system, and the output variable set is obtained from the evaluation results; a three-layer Bayesian causal network structure is constructed according to the endogenous variable set, exogenous input variable set and output variable set, and the conditional The independence test finds the causal relationship between variables; conducts system dynamics modeling and simulation calculations to obtain the equilibrium state of each variable according to the Bayesian causal network structure and the causal relationship between variables; maps the equilibrium state of each variable to the evaluation Indexes, the equilibrium state of each evaluation index under the constraints of exogenous conditions is obtained. The invention has the following advantages: it can effectively discover the causal relationship between the evaluation indexes in the complex system and obtain the equilibrium state of the evaluation indexes under different conditions.

The invention discloses a salient image extraction processing method and system. The method comprises the steps: analyzing the characteristics of the current image through an RGB channel so as to obtain the RGB salient characteristics of the current image; analyzing the characteristics of the current image through the Depth channel so as to obtain the Depth salient characteristics of the current image; both the RGB salient characteristics and the Depth salient characteristics meet conditional independent distribution and are assumed to obey Gaussian distribution; and performing salient characteristic fusion based on the Bayesian framework to estimate the saliency posterior probability so as to obtain the image saliency area. Therefore, the invention has the beneficial effects that: the high-level saliency characteristics of the RGB image and the Depth image are extracted by using the deep convolutional neural network, the correlation of the saliency characteristics is analyzed, the saliency characteristics are fused under the Bayesian framework and 3D saliency detection is modeled by using the DMNB generation model so as to obtain good accuracy, recall rate and F-measure.

A method for predicting sensor output values of a machine sensor monitoring system includes providing a set of input sensor data X and a set of output sensor data Y for a plurality of sensors the monitor the performance of a machine, learning a functional relationship that maps the input sensor data to the output sensor data by maximizing a logarithm of a marginalized conditional probability function P(Y|X) where a dependence of the output sensor data Y with respect to unknown hidden machine inputs u has been marginalized, providing another set of input sensor data X′, and calculating expected values of the output sensor data Y′ using the input sensor data X′ and the marginalized conditional probability function P(Y|X′), where the calculated expectation values reflect the dependence of the output sensor data Y″ with respect to the unknown hidden machine inputs u.

The invention discloses an electric power information network fault locating method. The relation between detection in a candidate detection set and nodes in an electric power information network is described by using a Bayesian network model. The information gain of single detection and the number of important nodes in a single detection path are combined to act as detection values to measure the diagnostic capacity of each detection in the candidate detection set. The detection of the maximum detection value is selected out of the candidate detection set to form a fault locating set. The most possible state information of the electric power information network is obtained by the returned detection result so as to locate fault nodes. The method is simple in theory, the fault locating process is enabled to be clearer by the directed edge between the information nodes and detection, and the new detection values are defined to act as the standard of selecting the fault locating set so that the accuracy of fault locating can be enhanced; and the Bayesian network is divided into multiple sub-networks by using the conditional independence of the Bayesian network, and the detection of the same sub-network is updated when the detection values are updated so that detection selection time can be reduced and the timeliness of fault locating can be enhanced.

The invention discloses an industrial alarming device design method based on dynamic fusion of a global uncertainty evidence, and belongs to the technical field of industrial alarming device designing. According to the method, collected procedure variable data is converted to alarming evidences of each moment based on a continuous type Sigmoid membership function and a Gaussian type membership function, a certain numerical value is given to global uncertainty information in the process, and processing on an uncertainty evidence can be better presented; linear update is conducted on rules by adopting a conditional evidence, updating fusion is conducted on the alarming evidence at the current moment and the global alarming evidence in the last moment, the global alarming evidence at the current moment is finally obtained, and alarming decision making is conducted according to related decision making principles. By means of the method, the random uncertainty and the recognition uncertainty of the procedure variable are effectively reduced in the updating fusion, and the accuracy of the alarming device is improved.

The invention provides a robustness prospect detection method based on multi-view learning. The method includes the steps that a reference background image is acquired by an input video through time domain median filtering method, and iterative search and multi-scale fusion are conducted on a current image and the reference background image to acquire heterogeneous characteristics; conditional probability density of prospects and conditional probability density of backgrounds are calculated by using condition independence of the heterogeneous characteristics, and a prospect posteriori probability and a background posteriori probability are calculated by using Bayes rules according to prospect likelihood, background likelihood and a prior probability; an energy function of a Markov random field model is established by means of the prospect posteriori probability, the background posteriori probability and space-time consistency constraint, the energy function is minimized by using a belief propagation algorithm, and segmented results of the prospect and the background are obtained. By means of the method, in a complex challenging environment, robustness prospect detection is achieved.

The invention discloses a weighted naive Bayes indoor positioning method based on attribute independence, and belongs to the technical field of indoor positioning, and the method comprises the following steps: building a CSI sample set of a position point; performing CSI data preprocessing; extracting main features through a PCA algorithm; establishing an offline fingerprint database; in the online stage, using a weighted naive Bayes positioning algorithm with independent attributes; in the offline stage, through multiple times of sampling analysis, knowing that CSI amplitude values of any position obey normal distribution, and therefore the mean value and the variance of the amplitude values of all the positions serve as fingerprints to be stored. In the online stage, the variance contribution rate calculated in the principal component analysis stage is used as a weight to be applied to naive Bayes classification, and the advantages of principal component analysis are maximized. According to the method, only the mean value and the variance of the CSI amplitude values measured by each reference point for multiple times need to be selected as fingerprints, the data is processed by using the principal component analysis method, the conditional independence assumption of the naive Bayes classifier is met, and the positioning precision is improved.

The invention discloses a short text classification method based on multiple weak supervision integration, and the method comprises the steps: obtaining an original data set and a knowledge base, andcarrying out the data preprocessing; carrying out knowledge extraction on the preprocessed data; representing the extracted knowledge as an annotation function, and using the annotation function for data annotation; carrying out label integration through a conditional independent model; training a classification model based on a full-connection neural network; evaluating and optimizing the classification model to obtain an optimal model; and performing short text classification by utilizing the optimal model. According to the short text classification method based on multiple weak supervisionintegration, explicit knowledge and implicit knowledge are completely expressed in a mode of combining keyword matching, regular expression and remote supervision clustering; by means of probability labels generated by a label integration mechanism, automatic labeling of label-free data is achieved, the problem of data sparsity of short texts is relieved, and the problem of unbalanced classification of the short texts is effectively solved.

The invention discloses a cross-modal data retrieval method and system based on graph regularization and modal independence, and the method comprises the steps of receiving the original data of different modalities, carrying out the feature extraction, building a multi-modal data set which consists of the image text pairs in one-to-one correspondence, wherein the multi-modal data set comprises a training set and a test set; projecting the feature matrixes of different modal data in the training set to a public subspace through optimizing a predefined objective function to obtain an image projection matrix and a text projection matrix; according to the image projection matrix and the text projection matrix, projecting the feature matrixes of different modal data in the test set to a publicsubspace; calculating the similarity between the projected matrix and other projection matrixes in the public subspace; and performing descending sort according to the similarity to obtain the data corresponding to the plurality of first feature projection matrixes, and performing the cross-mode retrieval.

The techniques disclosed herein determine a location independent network delay, via a network monitoring device. Such techniques particularly include determining various delays caused by retransmission in instances of lost data segments and application response time.

The invention discloses a calculating method of the imprecise probability of the steady-state availability of an electric power system. In the prior art, the probability interval of the steady-state availability of the electric power system is deducted according to interval reliability indexes of an element by employing an interval operation or optimization algorithm, the calculating amount is large, and the engineering realization is very difficult. According to the method, imprecise probability deduction of condition distributions of the state staying time of the system is performed by employing a gamma exponential model so that interval values expected by the state staying time conditions can be obtained; and by employing the Markov property of the electric power system, namely the conditional independence of the state staying time of the electric power system, upper and lower bound expressions of the intervals of the steady-state availability of the electric power system are deducted by directly employing sample data of the system without calculating the imprecise reliability indexes of the element.

The invention discloses a municipal facility fault prediction method and device. The method comprises the following steps: acquiring data of a plurality of state classification features of municipal facilities collected in real time, pretreatment, determining whether the plurality of state classification features satisfy a condition independence hypothesis and / or a weight equal hypothesis, if not,optimizing a preset naive Bayesian classifier through the Pearson correlation coefficient and / or entropy method according to the plurality of state classification features; and inputting the preprocessed data into an optimized naive Bayesian classifier to obtain the fault prediction probability of the municipal facilities, and determining the fault prediction of the municipal facilities accordingto the sequence of the fault prediction probability of the municipal facilities in all the fault prediction probabilities of the municipal facilities. The problems that at present, municipal facilitymaintenance is low in efficiency and high in cost, and requirements cannot be matched in time due to the fact that more citizens actively report repair and conduct manual inspection can be solved.

The invention discloses a Laos language text subject classification method, belonging to the technical field of natural language processing and machine learning. The N-gram language feature extractionmodel and naive bayesian mathematical model to achieve the recognition of laos article theme, to some extent, eliminated the limitations of naive bayesian. It considers the conditional independence assumption that the text is regarded as a word bag model without considering the order information between words, and simultaneously uses the unigram and bigram feature model, which improves the recognition rate of the text.

The invention discloses a method for predicting mineral resources by an evidence weight model, which includes such steps as adding evidential layer to the model in order of significance to predict object, obtaining obvious effect, and carrying out the prediction result for further optimizing and evaluating predicted area with very important reference significance. Finally, according to the posterior probability map obtained by the improved evidence weight method, the favorable ore-forming areas are delineated, and the prediction results have an important guiding role in the next step of ore prospecting deployment. The invention can reduce the influence on the prediction result due to the fact that the evidence layer does not satisfy the conditional independence assumption.

The invention discloses a method and system for phase position unwrapping based on Markov random field (MRF). The method includes the steps that according to features of real phase positions, through MRF, modeling is carried out on the variable space formed by the real phase positions and observation phase positions; condition independence among all nodes on the MRF is analyzed, and according to the condition independence, the relation between the total joint probability distribution of the MRF and potential energy of the MRF is analyzed; according to the relation and the features of the real phase positions, the potential energy of the MRF is set and minimized to obtain estimated values of the real phase positions. The real phase positions can be recovered from the observation phase positions better.

Systems and methods for predictive network modeling are disclosed. The systems and methods disclosed compute a top-down causal model and a bottom-up predictive model and utilize those models to determine the conditional independence among multiple variables and causality among equivalent variable structures. Before or during modeling, the data is passed through Markov Chain Monte Carlo sampling.

The embodiment of the invention discloses a classifying method based on a multi-label double-view support vector machine, which comprises the steps of first, defining a novel distance measuring method in a multi-label space to measure the distance from point to point in the multi-label space under a special classifying target; then, extracting two groups of feathers of a training set from two conditionally independent angles of view, and combining and utilizing complementary information of the two groups of feathers comprised by the two angles of view; and finally, by combining information in the multi-label space and the double-view space, carrying out multi-label classifying training by using the defined novel multi-label double-view support vector machine. The classifying method based on the multi-label double-view support vector machine is used to handle the multi-label classifying problem by an identifying classifier which combines and utilizes information comprised in the multi-label space and information in the multi-view angles, and the noise of the labels in the training set is reduced while a more accurate classifying method is obtained.

The invention discloses a novel modelless Bayesian classification and prediction model soft measurement method. Firstly, the dimension reduction and the noise reduction of the gas chromatogram data are effectively realized through the curve fitting method, and then the characteristic value of the gas chromatogram data is extracted, thereby shortening the classification model training time and getting better generalization ability. The novel modelless Bayesian classification and prediction model soft measurement method uses a new modelless Bayesian classification algorithm to establish a recognition model, which can effectively avoid the problem of the decline of the generalization performance of the model caused by the training sample not satisfying the condition independence. The novel modelless Bayesian classification and prediction model soft measurement method provided by the invention objectively shows the degree of flooding of oil and gas reservoirs under different conditions through gas chromatogram measurement, and indicates the degree of flooding and the exploitation value of each oil and gas reservoir, thereby helping oil drilling companies to further improve mining efficiency and reduce costs. Therefore, the technical scheme provided by the present invention has the validity and applicability.

The invention discloses a method for constructing a time sequence causal relationship graph, and the method comprises the steps: calculating a first time lag value of each time sequence and a second time lag value between every two time sequences based on a plurality of time sequences; a direct lag dependent variable and an initial connection graph for each time sequence are determined. And judging whether a causal relationship exists between the time sequences corresponding to every two mutually connected nodes in the initial connection graph or not by utilizing a conditional independence criterion, thereby obtaining an intermediate connection graph. And after determining the direction of the undirected edge between the time sequences with the causal lagging relationship in the intermediate connection graph, checking whether the undirected edge between every two time sequences at the current moment actually exists or not by using the conditional independence criterion again, and obtaining a final time sequence causal relationship graph. According to the scheme, every two time sequences are fitted through the first time lag value to obtain the residual sequence, and the second time lag value is calculated by using the residual sequence, so that the accuracy of causal relationship judgment can be improved.