45 results about "Probabilistic inference" patented technology

A method of estimating the navigational state of a system entails acquiring observation data produced by noisy measurement sensors and providing a probabilistic inference system to combine the observation data with prediction values of the system state space model to estimate the navigational state of the system. The probabilistic inference system is implemented to include a realization of a Gaussian approximate random variable propagation technique performing deterministic sampling without analytic derivative calculations. This technique achieves for the navigational state of the system an estimation accuracy that is greater than that achievable with an extended Kalman filter-based probabilistic inference system.

A method of estimating the navigational state of a system entails acquiring observation data produced by noisy measurement sensors and providing a probabilistic inference system to combine the observation data with prediction values of the system state space model to estimate the navigational state of the system. The probabilistic inference system is implemented to include a realization of a Gaussian approximate random variable propagation technique performing deterministic sampling without analytic derivative calculations. This technique achieves for the navigational state of the system an estimation accuracy that is greater than that achievable with an extended Kalman filter-based probabilistic inference system.

A system and method for multi-modal context-sensitive applications in a home network are provided. The system includes an input processing agent classifying inputs from a predetermined input device based on at least one criterion selected from the group consisting of an input block, an input type, a priority, and ranking and then outputting the classified inputs; an input collaboration agent performing real-time analysis on each of the inputs fed by the input processing agent to obtain a pre-defined output action for an input type corresponding to each input based on a user's context and performing a probabilistic inference on the inputs fed by the input processing agent to obtain a possible output action based on the user's context; and an agent server obtaining an output resource, which is available in the home network and corresponds to the user's context, based on the output action obtained by the input collaboration agent.

The computational efficiency of Finite Element Methods (FEM) on parallel architectures is typically severely limited by sparse iterative solvers. Standard iterative solvers are based on sequential steps of global algebraic operations, which limit their parallel efficiency, and prior art techniques exploit sophisticated programming techniques tailored to specific CPU architectures to improve performance. The inventors present a FEM Multigrid Gaussian Belief Propagation (FMGaBP) technique that eliminates global algebraic operations and sparse data-structures based upon reformulating the variational FEM into a probabilistic inference problem based upon graphical models. Further, the inventors present new formulations for FMGaBP, which further enhance its computation and communication complexities where the parallel features of FMGaBP are leveraged to multicore architectures.

A system and method for multi-modal context-sensitive applications in a home network are provided. The system includes an input processing agent classifying inputs from a predetermined input device based on at least one criterion selected from the group consisting of an input block, an input type, a priority, and ranking and then outputting the classified inputs; an input collaboration agent performing real-time analysis on each of the inputs fed by the input processing agent to obtain a pre-defined output action for an input type corresponding to each input based on a user's context and performing a probabilistic inference on the inputs fed by the input processing agent to obtain a possible output action based on the user's context; and an agent server obtaining an output resource, which is available in the home network and corresponds to the user's context, based on the output action obtained by the input collaboration agent.

A fast variational on-line learning technique for training a transformed hidden Markov model. A simplified general model and an associated estimation algorithm is provided for modeling visual data such as a video sequence. Specifically, once the model has been initialized, an expectation-maximization (“EM”) algorithm is used to learn the one or more object class models, so that the video sequence has high marginal probability under the model. In the expectation step (the “E-Step”), the model parameters are assumed to be correct, and for an input image, probabilistic inference is used to fill in the values of the unobserved or hidden variables, e.g., the object class and appearance. In one embodiment of the invention, a Viterbi algorithm and a latent image is employed for this purpose. In the maximization step (the “M-Step”), the model parameters are adjusted using the values of the unobserved variables calculated in the previous E-step. Instead of using batch processing typically used in EM processing, the system and method according to the invention employs an on-line algorithm that passes through the data only once and which introduces new classes as the new data is observed is proposed. By parameter estimation and inference in the model, visual data is segmented into components which facilitates sophisticated applications in video or image editing, such as, for example, object removal or insertion, tracking and visual surveillance, video browsing, photo organization, video compositing, and meta data creation.

A computer-implemented method of querying a source dataset in which a user provides a query to a dataset querying system. The system automatically processes both the dataset and the query, so that processing the query influences the processing of the dataset, and / or processing the dataset influences the processing of the query. The system automatically processes the query and the dataset to derive a probabilistic inference of the intent behind the query. The user interacts with the relevance-ranked attempts to answer that query and the system then iteratively improves or varies how it initially processed the query and the dataset, to dynamically generate and display further relevance-ranked attempts to answer that query, to enable the user to iteratively explore the dataset or reach a useful answer.

A system and method to integrate breadth-first and depth-first strategies in a single search technique or routine is provided. It combines the complementary strengths of both strategies to achieve significantly improved speed over either strategy used alone. The new algorithm can be used to efficiently find solutions to the treewidth problem that has applications in areas such as diagnosis using probabilistic inferences.

Analysis evaluates formation fluid with a downhole tool disposed in a borehole. A plurality of possible constituents is defined for the formation fluid, and constraints are defined for the possible constituents. The constraints can include boundary constraints and constraints on the system's dynamics. The formation fluid is obtained from the borehole with the downhole tool over a plurality of time intervals, and density of the obtained formation fluid is obtained at the time intervals. To evaluate the fluid composition, a state probability distribution of the possible constituents of the obtained formation fluid at the current time interval is computed recursively from that at the previous time interval and by assimilating the current measured density of the obtained formation fluid in addition to the defined boundary / dynamic constraints. The probabilistic characterization of the state of the possible constituents allows, in turn, the probabilistic inference of formation properties such as contamination level and GOR.

According to one embodiment, a method comprises evaluating messages between nodes of a distributed computing environment. Based on timing relationships of the messages, probability of causal links between pairs of messages determined. Based on the determined probability, at least one causal path comprising a plurality of causal links is inferred. Further, an overall probability that the inferred causal path is accurate is determined. In certain embodiments, the overall probability is output for the corresponding causal path to provide an indication of the confidence of the accuracy of such inferred causal path.

The subject invention leverages standard probabilistic inference techniques to determine a log-likelihood for a conditional Gaussian graphical model of a data set with at least one continuous variable and with data not observed for at least one of the variables. This provides an efficient means to compute gradients for CG models with continuous variables and incomplete data observations. The subject invention allows gradient-based optimization processes to employ gradients to iteratively adapt parameters of models in order to improve incomplete data log-likelihoods and identify maximum likelihood estimates (MLE) and / or local maxima of the incomplete data log-likelihoods. Conditional Gaussian local gradients along with conditional multinomial local gradients determined by the subject invention can be utilized to facilitate in providing parameter gradients for full conditional Gaussian models.

A scalable hierarchical coreference method that employs a homomorphic compression scheme that supports addition and partial subtraction to more efficiently represent the data and the evolving intermediate results of probabilistic inference. The method may encode the features underlying conditional random field models of coreference resolution so that cosine similarities can be efficiently computed. The method may be applied to compressing features and intermediate inference results for conditional random fields. The method may allow compressed representations to be added and subtracted in a way that preserves the cosine similarities.

A fast variational on-line learning technique for training a transformed hidden Markov model. A simplified general model and an associated estimation algorithm is provided for modeling visual data such as a video sequence. Specifically, once the model has been initialized, an expectation-maximization (“EM”) algorithm is used to learn the one or more object class models, so that the video sequence has high marginal probability under the model. In the expectation step (the “E-Step”), the model parameters are assumed to be correct, and for an input image, probabilistic inference is used to fill in the values of the unobserved or hidden variables, e.g., the object class and appearance. In one embodiment of the invention, a Viterbi algorithm and a latent image is employed for this purpose. In the maximization step (the “M-Step”), the model parameters are adjusted using the values of the unobserved variables calculated in the previous E-step.

The present invention relates to a system and a method of making optimal medical decisions. In one embodiment presented for illustration the system comprises a quantitative model of the disease in the form of transition probabilities, the quantitative model of the effect of the medical treatment (therapy, drug or remedy) on the course of the disease, the quantitative model of costs and benefits, including monetary as well as non-monetary costs and benefits, and the model of preferences with respect to the costs and benefits. Using probabilistic inference, distributions of parameters of models are extracted from the data and the opinions of parties involved in the medical treatment. An expectation of the value of the treatment is computed. Optimality of the treatment is achieved by choosing the treatment or its parameters that give the greatest value given the evidence, the models, and the preferences. Other embodiments are discussed.

A system and a method are disclosed for simulating data driven market order exchange mechanics. An event processing engine receives a feed of market data and forwards it to an exchange simulator. The feed of market data may be recorded market data, live relayed market data, or simulated market data. A series of order requests is also received. The order requests are market order or limit orders, and can include new orders, amend orders, or cancel orders. The feed of market data is analyzed and an inference algorithm is applied by making probabilistic inferences to determine what actions may have occurred to produce the received feed of market data. A second series of order requests are produced. The received order requests and the second series of order requests are combined with normal exchange rules to produce a stream of simulated market data and a series of updated order requests.

A system and method to integrate breadth-first and depth-first strategies in a single search technique or routine is provided. It combines the complementary strengths of both strategies to achieve significantly improved speed over either strategy used alone. The new algorithm can be used to efficiently find solutions to the treewidth problem that has applications in areas such as diagnosis using probabilistic inferences.

A probability of an observation sequence stored at a client is evaluated securely with respect to a hidden Markov model (HMM) stored at a server. The server determines, for each state of the HMM, an encryption of a log-probability of a current element of the observation sequence. Determines, for each state of the HMM, an encryption of a log-summation of a product of a likelihood of the observation sequence based on a previous element of the observation sequence and a transition probability to the state of the HMM. Determines an encryption of a log-likelihood of the observation sequence for each state as a product of the encryption of a log-summation and an encryption of a corresponding log-probability of the current element of the observation sequence; and determines an encryption of the log-probability of the observation sequence based on the log-likelihood of the observation sequence for each state.

The subject invention leverages standard probabilistic inference techniques to determine a log-likelihood for a conditional Gaussian graphical model of a data set with at least one continuous variable and with data not observed for at least one of the variables. This provides an efficient means to compute gradients for CG models with continuous variables and incomplete data observations. The subject invention allows gradient-based optimization processes to employ gradients to iteratively adapt parameters of models in order to improve incomplete data log-likelihoods and identify maximum likelihood estimates (MLE) and / or local maxima of the incomplete data log-likelihoods. Conditional Gaussian local gradients along with conditional multinomial local gradients determined by the subject invention can be utilized to facilitate in providing parameter gradients for full conditional Gaussian models.

Parameters of a hidden Markov model (HMM) are determined by a server based on an observation sequence stored at a client, wherein the client has a decryption key and an encryption key of an additively homomorphic cryptosystem, and the server has only the encryption key. The server initializes parameters of the HMM and updates the parameters iteratively until a difference between a probability of the observation sequence of a current iteration and a probability of the observation sequence of a previous iteration is above a threshold, wherein, for each iteration, the parameters are updated based on an encrypted conditional joint probability of each pair of states given the observation sequence and the parameters of the HMM, wherein the encrypted conditional probability is determining in an encrypted domain using a secure multiparty computation (SMC) between the server and the client.

The invention discloses an indoor semantic trajectory labeling and completion method in a low sampling and positioning environment. The method includes the following steps: inputting semantic entity and indoor spatial structure information to construct the mobile transfer map; collecting the original position trajectory and dividing each original position trajectory into a plurality of segmented segments; for the event mode, the user selecting the bound position trajectory segment; comparing the similarities of the key features of the segmented fragments and the location trajectory fragments,and linking the most similar fragments to form a group of independent semantic tuples; assembling the semantic tuples, and generating the missing semantic tuples by probabilistic inference; exportingthe complete semantic trajectory after insertion and completion. The method of the invention can derive accurate and available complete semantic trajectory data when facing the indoor positioning scene with complex movement constraint and low sampling frequency of position data.

A demographic attribute value of a sink online document (such as Websites or Web pages) may be determined given a set of users who have visited at least one of the source documents and the sink document, by (a) accepting a value(s) of the demographic attribute, each of which values is associated with a source online document (where each of the source online documents has a value for the demographic attribute and has been visited by at least one user of the given set), (b) determining an estimate of the demographic attribute value of each of the users of the given set using the accepted demographic attribute value of each of the source online documents visited by the user, and (c) determining the demographic attribute value of the sink online document using the determined estimate of the demographic attribute value of each of the users of the given set.

[Problem] An objective of the present invention is to provide an analysis system whereby analysis of customer data which is collected and accumulated in a management company system is carried out in a short time with high precision. [Solution] Provided is a customer data analysis system (1) for analyzing itemized data which is collected and accumulated in a management company system (2), said customer data analysis system (1) comprising a predict processing means for treating a plurality of arbitrary items of the itemized data as explanatory variables, treating another plurality of items as target variables, and computing probability values of the plurality of items from correlations among the items by probabilistic inference.

A demographic attribute value of a sink online document may be determined by (a) accepting a value(s) of the demographic attribute of a source online document, (b) accepting, for each of the source online documents having an accepted demographic attribute value, a probability that a user will visit or has visited the sink online document if the user visited the source online document, and (c) determining the demographic attribute value of the sink online document using the demographic attribute value of each of the source online documents and the probabilities. A demographic attribute value of the sink online document may also be determined using the above information and using the demographic attribute value of each of the source online documents and the probabilities that a user will visit or has visited the sink online document if the user visited the other online document.

A most likely sequence of states corresponding to an observation sequence stored at a client is determined securely with respect to a HMM stored at a server. An encryption of a log-probability of the current element of the observation sequence is determined for each state of the HMM. A product of an encryption of the log-probability of the state for the current element, an encryption of a transition probability to the state, and the encryption of a log-probability of the current element of the observation sequence is determined iteratively, for each state of the HMM, to produce an encrypted matrix of indexes of the states; and the encrypted matrix is transmitted to the client.

A segmentation free method and system for automatic license plate recognition. An OCR classifier can be swept across an image of a license plate. Characters and their locations can be inferred with respect to the image of the license plate using probabilistic inference based on a Hidden Markov Model (HMM). A language model can be combined with a license plate candidate from the HMM to infer the optimal or best license plate code. The language model can be configured by employing a corpus of license plate codes, wherein the corpus includes a distribution representative of training sets and tests sets.

A probability of an observation sequence stored at a client is evaluated securely with respect to a hidden Markov model (HMM) stored at a server. The server determines, for each state of the HMM, an encryption of a log-probability of a current element of the observation sequence. Determines, for each state of the HMM, an encryption of a log-summation of a product of a likelihood of the observation sequence based on a previous element of the observation sequence and a transition probability to the state of the HMM. Determines an encryption of a log-likelihood of the observation sequence for each state as a product of the encryption of a log-summation and an encryption of a corresponding log-probability of the current element of the observation sequence; and determines an encryption of the log-probability of the observation sequence based on the log-likelihood of the observation sequence for each state.

A demographic attribute value of a sink online document (such as Websites or Web pages) may be determined given a set of users who have visited at least one of the source documents and the sink document, by (a) accepting a value(s) of the demographic attribute, each of which values is associated with a source online document (where each of the source online documents has a value for the demographic attribute and has been visited by at least one user of the given set), (b) determining an estimate of the demographic attribute value of each of the users of the given set using the accepted demographic attribute value of each of the source online documents visited by the user, and (c) determining the demographic attribute value of the sink online document using the determined estimate of the demographic attribute value of each of the users of the given set.

Aspects of the present invention provide systems and methods for distributed signal processing of indoor localization signals wherein statistical algorithms and machine learning are used in place of a fingerprint map. The disclosure relates to calculation of angle and distance based on measurements of an indoor localization signal, followed by energy-efficient distribution of signal processing. Local signal processing is performed using any of multiple eigen structure algorithms or a linear probabilistic inference, before cloud-based signal processing is performed using a nonlinear probabilistic inference and machine learning that's been trained with historical data transmitted by the base stations and time-of-day location patterns. Without having to generate and constantly update an energy-exorbitant fingerprint map, the disclosed system reduces localization error to merely 50 cm with 95% probability without compromising energy-efficiency to rival the accuracy of indoor localization systems that utilize fingerprinting.

The present disclosure relates to business data processing and facilitates fusing business data spanning disparate sources for processing distributional queries for enterprise business intelligence application. Particularly, the method comprises defining a Bayesian network based on one or more attributes associated with raw data spanning a plurality of disparate sources; pre-processing the raw data based on the Bayesian network to compute conditional probabilities therein as parameters; joining the one or more attributes in the raw data using the conditional probabilities; and executing probabilistic inference from a database of the parameters by employing an SQL engine. The Bayesian Network may be validated based on estimation error computed by comparing results of processing a set of validation queries on the raw data and the Bayesian Network.