32 results about "Bayesian risk" patented technology

A hypothesis space of a search graph may be determined. The hypothesis space may include n hypothesis-space transcriptions of an utterance, each selected from a search graph that includes t>n transcriptions of the utterance. An evidence space of the search graph may also be determined. The evidence space may include m evidence-space transcriptions of the utterance that are randomly selected from the search graph, where t>m. For each particular hypothesis-space transcription in the hypothesis space, an expected word error rate may be calculated by comparing the particular hypothesis-space transcription to each of the evidence-space transcriptions. Based on the expected word error rates, a lowest expected word error rate may be obtained, and the particular hypothesis-space transcription that is associated with the lowest expected word error rate may be provided.

The invention discloses a head-shoulder sequence image segmentation method based on double-pattern matching and edge thinning, and the method is carried out as per the following steps: step 1, determining a face position, namely, adopting the Bayesian risk decision to determine the face area by considering the distribution of the face on a color plane; step 2, determining the head-shoulder area, namely, considering the head-shoulder area as a combination of two rectangular areas, one as head rectangle and the other as shoulder rectangle, setting the width of the shoulder rectangle as three times of the width of the head rectangle, and finally determining the head rectangle through adopting two rectangular moving templates and taking the proportion falling in the two template areas as the matching standard; and step 3, edge thinning, namely, adopting the Canny edge detection operator to acquire the accurate contour of a moving object. The head-shoulder sequence image segmentation method has the benefits that the Bayesian risk decision mechanism is adopted to determine the face position, the double-pattern matching algorithm is adopted to further determine the head-shoulder area, and finally the edge thinning is conducted, as a result, the algorithm can efficiently segment the head-shoulder sequence.

The invention provides a cost-sensitive incremental face recognition method based on information entropy selection. The cost-sensitive incremental face recognition method is composed of a deep convolutional neural network part, an information entropy-based sample selection part and a cost-sensitive sequential three-branch decision classification part. The information entropy is used for assessingthe information amount of the classification result of the face recognition sample, so that the system can automatically assess the unlabeled sample information amount, and samples with large information amount are selected for manual labeling; a face recognition problem is regarded as a sequential process of information granularity from coarse to fine by utilizing the thought of three sequentialdecisions sensitive to cost, each iterative loop added with a marked sample is used as a decision step of the three sequential decisions, and the minimum cost recognition effect of the sample in eachdecision step is given according to the minimum Bayesian risk principle.

The invention discloses a high-precision face recognition method for a complex face recognition access control system, belonging to the field of face image processing. The method comprises the following steps: face vector normalization, cost-sensitive face feature dimension reduction and model prediction. The method adopts the cost-sensitive face feature dimension reduction to give different misclassification cost for the different kinds of misclassification in the access control system. The extracted features meet the smallest bayesian risk criteria, the discriminant ability of the extracted features is increased, and further the accuracy of recognition is increased. In addition, the cost-sensitive face feature dimension reduction has robustness to the noises existing in the access control system, imbalanced categories and different kinds of data distribution density, and the method has a higher use value.

The invention provides a SKIP mode quickly selecting method based on Bayesian minimum hazard decision. The SKIP mode quickly selecting method based on Bayesian minimum hazard decision comprises the concrete steps of taking the rate-distortion cost of the SKIP mode of a current coding unit as decision features to count and learn first four frames, and for different quantization parameters and different coding unit depths, performing non-parametric probability density estimation to obtain the conditional probability distribution of the decision features in an optimal SKIP mode and an optimal non-SKIP mode; based on the statistical information of online learning, utilizing a Bayesian minimum hazard model to discriminate the SKIP mode of surplus frames in advance. The SKIP mode quickly selecting method based on Bayesian minimum hazard decision can rapidly achieve SKIP mode discrimination under the condition of ensuring the coding performance, thereby skipping predictive coding mode computation, effectively reducing the coding complexity of an HEVC (high efficiency video coding) encoder and facilitating implementation of real-time application of the HEVC encoder.

The invention relates to the field of water conservancy and hydropower engineering, and discloses a Bayesian network model used for cascade hydropower stations risk analysis and a construction method. Various risk factors of a cascade hydropower station are comprehensively considered. Strong theoretical and technical supports are provided for hydropower engineering risk identification and assessment. The construction method provided by the invention comprises the steps that a an engineering risk decomposition identification method based on the basic functions of safe operation is used to identify the function affecting risk of the cascade hydropower station; b all specific risk factors which affect the cascade hydropower station are comprehensively considered, and are defined as node variables to form a variable set; c the risk factors in the variable set are hierarchically filled in a table from high to low, so as to acquire a risk factor causal logic relationship hierarchy table; and d according to the causal relationship among the variables in the risk factor causal logic relationship hierarchy table, a single arrow is used to connect all variables into a directed acyclic graph to complete the construction of the Bayesian risk network model used for the cascade hydropower station.

The embodiment of the invention provides a cattle illness data analysis method and device. The method concretely comprises the following steps: obtaining state index data representing cattle health states; selecting sample data, performing sample training of selected sample data according to an initialized cost matrix based on illness classification in advance and a multi-classification Bayesian process classifier based on cost sensibility, and obtaining the illness classification model of training samples; and according to the obtained state index data, the cost matrix and the training model, calculating the illness classification probability of the cattle to be detected and outputting the illness classification probability of the cattle to be detected. The method provided by the embodiment of the invention provides a cost-sensitive multi-classification Bayesian and Gaussian process to process the calculation of the illness classification probability of cattle, perform calculation of the state index states of the cattle and output a plurality of illness classification probabilities, and misjudgement cost factors are added to overcome the defects that a current discriminating method only can obtain general low misjudgement rate and only can output single classification results.

The invention discloses a key risk factor and key risk transfer path identification method of cascade hydropower stations based on the Bayesian risk network, and relates to the field of hydraulic andhydroelectric engineering. The risk of the cascade hydropower stations is characterized by a common-cause failure problem, the advantage of the Bayesian network in solving the common-cause failure problem is utilized, and disadvantages of the traditional fault, event and decision trees in the aspect of risk analysis of hydroelectric engineering are overcome. On the basis of a Bayesian risk networkmodel of the cascade hydropower station, the key risk factor and key risk transfer path identification method is provided, and theoretical and method support for using the Bayesian network method ofthe cascade hydropower stations in risk identification and prevention is provided. The method is simple, clear and highly operable, and has significant application prospects in using the Bayesian network for risk identification and prevention in systems engineering as cascade hydroelectric engineering and civil engineering.

The invention discloses an ultrasonic phased array method based on information fusion based on information fusion and belongs to an ultrasonic signal processing technology. The algorithm comprises a bitmap segmenting method and bayes estimation. The bitmap segmenting method comprises the following steps: extracting out medians, upper quartiles and lower quartiles of signals independently collected by each array element of an ultrasonic phased array probe in sequence, obtaining quartile dispersion, solving out elimination points and obtaining effective data fusion sets. Bayes estimation is established by virtue of the following steps: obtaining a characteristic function of each piece of effective data, solving out a decision function, obtaining a risk function, extracting the bayes risk of the decision function and obtaining most effective estimated values of parameters. The method disclosed by the invention can be used for improving the utilization rate of information collected by ultrasonic phased array elements, and obtaining more accurate and reliable description on a real environment.

The invention relates to an image noise reduction method of Contourlet transformation region, wherein it comprises: translating the input image with noise; using Contourlet transformation to decompose said image at multiple sizes and directions, and in the Contourlet transformation region, using minimum Bayesian risk function to evaluate the Contourlet region factor; then processing Contourlet inverse transformation and inversed translating to obtain the noise reduced image; then repeating aforementioned steps, linearly averaging the noise reduced image, to obtain last noise reduced image. The invention can improve the quality of noise reduced image, to provide full and accurate target and background information, to be used in optical imaging, target detecting and safety detecting systems.

The invention provides a multi-sensor damage networking monitoring method based on Bayesian risk function. The method includes the steps of firstly, selecting appropriate sensors; secondly, building the Bayesian risk function; thirdly, building a Bayesian risk function optimization equation, and searching for an optimal layout scheme; fourthly, performing sensor layout scheme quantitative analysis, and using detection rate and false alarm rate to quantify the advantages and disadvantages of the given sensor layout scheme; fifthly, performing sensor layout damage networking monitoring. By the arrangement, the method has the advantages that multi-sensor damage networking monitoring based on the Bayesian risk function is achieved, complex-structure damage signal effective extraction is achieved, and influence of complex-structure dimension and configuration on monitoring signal transmission and collection is lowered; limited resource distribution is optimized, system monitoring ability isincreased, and equipment whole life cycle cost is lowered; the multi-filed detection method is adopted, and the problems that the piezoelectric technology is greatly affected by the structure, the optical fiber technology cannot directly monitor cracks, and intelligent coating is high in false alarm rate are solved.

The invention relates to the field of water conservancy and hydropower engineering, and discloses a construction method of Bayesian risk network models of a cascade reservoir group. Risk transmission factors of the cascade reservoir group are comprehensively considered, and powerful theoretical and technical support is provided for risk analysis and evaluation of the cascade reservoir group. The method comprises the following steps that firstly, cascade hydropower station Bayesian risk network models of different dam types are generalized by selecting high-level network nodes; secondly, a reservoir inflow flood node and a reservoir outflow flood node on the generalized cascade hydropower station Bayesian risk network model are set to be connection nodes between an upstream hydropower station and a downstream hydropower station which are adjacent to each other; thirdly, the generalized cascade hydropower station Bayesian risk network model serves as a basic unit, the reservoir outflow flood node of an upstream reservoir and the reservoir inflow flood node of a downstream reservoir are connected through a single arrow line according to the combination modes between the cascade reservoirs, and the Bayesian risk network models of the cascade reservoir group in different combination modes are formed.

Methods are provided to obtain task allocation, planning, and scheduling necessary to ensure service agents can accomplish their tasks efficiently, while transport agents perform the necessary transport and refueling operations. The methods take a scheduling-centric approach, formally incorporate fuel constraints, and are generalized for an arbitrary number of service agents and transport agents. The methods provide for the formal definition of the novel simple agent transport problem, the unique modeling and constraint-based logic required for the docking, transport, and deployment of service agents and the resulting analysis. Tasks to be completed consist of a set of locations, and the agents must service each location set. The methods and systems incorporate Bayes risk to schedule slips and to the application of service agent-transport agent scheduling.

The invention provides a multi-sensor multi-target joint detection, tracking and classification method, which is characterized in that the method comprises the following steps: S1, defining a new Bayesian risk given the initial value of the multi-target state; S2, predicting the multi-objective state under the condition of the category hypothesis in the category hypothesis set to obtain the priorstate distribution of the multi-objective; 3, under the condition of a class decision set, calculating a multi-objective posterior density at the time k to obtain a posterior multi-objective state distribution; S4, calculating multi-target detection loss, state estimation loss and classification loss under different decision conditions; S5: according to the detection loss, the state estimation loss and the classification loss, the multi-objective estimation and the classification optimal solution are obtained based on the minimum Bayesian risk criterion. This method is easy to implement and provides important technical support for multi-sensor networking environment sensing system.

An ultrasonic phased array nondestructive test method based on fractal is disclosed. The method comprises following two parts: Bayesian information fusion algorithm and signal fractal processing. The Bayesian information fusion algorithm includes following steps of: taking each array element of a ultrasonic phased array probe, separately collecting a median, an upper quartile and a lower quartile in signals in sequence to obtain a quartile dispersion, solving an eliminative point to obtain an effective data fusion set, solving characteristic functions and decision functions of effective data to obtain a risk function, and extracting Bayesian risk of the decision functions to obtain optimal estimated values of parameters. The signal fractal processing includes following steps of: analyzing image fractal characteristics through a wavelet decomposition coefficient figure to determine a scale-free zone, calculating a fractal dimension, and determining whether a flaw exists and determining a flaw type according to the fractal dimension. Based on adoption of the Bayesian information fusion algorithm to improve the information utilization rate, the method adopts a fractal technology to process nonlinear modern signals so as to determine whether a flaw exists in a composite material and to determine a flaw type.

The invention provides a digital fire safety assessment and visual operation and maintenance method, which meets the requirements of fire monitoring and early warning, and is characterized by comprising the following steps: S1, formulating a risk level and a quantitative range according to a Bayesian risk identification and analysis model; s2, environment index monitoring; wherein the monitoring comprises cable residual current monitoring, cable temperature monitoring or cracking quality monitoring, and when a corresponding index is monitored to exceed a preset standard, an early warning is initiated; s3, open fire monitoring and alarming; the multiple cameras simultaneously take frames and send pictures to the AI image recognition model, and the AI image recognition model carries out recognition and risk grading on open fire or smoke and fire in the pictures and feeds back a processing result to the monitoring background; when the AI image recognition model recognizes open fire, an adjacent fire-fighting spraying device is called according to the coordinates of the corresponding camera to carry out spraying and fire extinguishing.

A computer-implemented method for linear classification involves generating a data-driven likelihood ratio test based on a dual locus of likelihoods and principal eigenaxis components that contains Bayes' likelihood ratio and automatically generates the best linear decision boundary. A dual locus of likelihoods and principal eigenaxis components, formed by a locus of weighted extreme points, satisfies fundamental statistical laws for a linear classification system in statistical equilibrium and is the basis of an optimal linear classification system for which the eigenenergy and the Bayes' risk are minimized, so that the classification system achieves Bayes' error rate and exhibits optimal generalization performance. Linear classification systems can be linked with other such systems to perform multiclass linear classification and to fuse feature vectors from different data sources. Linear classification systems also provide a practical statistical gauge that measures data distribution overlap and Bayes' error rate.

The invention discloses a text generation method and device, electronic equipment and a medium. The method comprises the steps that a text candidate set is obtained, the text candidate set comprises a plurality of candidate texts, and each candidate text is a text obtained after a to-be-processed text is processed; performing at least two times of screening method processing on the text candidate set by referring to a screening strategy of minimum Bayesian risk decoding to obtain a target text; wherein the screening methods are screening methods with different functions, the output of each screening is used as the input of the next screening method, the result of each screening output is determined based on the evaluation score of the screened candidate text, and the evaluation score of the screened candidate text is determined based on the evaluation index of the used screening method. According to the method, the obtained text candidate set is processed through the screening method at least twice, and the problem that the result generated by a given text generation model is poor can be solved.

The invention provides a physical layer spoofing detection method based on deep reinforcement learning, and mainly solves the problems that a dynamic unknown wireless environment, a channel model or parameters are difficult to obtain and a fixed detection threshold is difficult to accurately select in the existing physical layer spoofing detection method. The method comprises the following implementation steps: 1) establishing a spoofing attack scene, and extracting physical layer channel information between a receiving party and a transmitting party by a receiving party to represent physical layer fingerprint characteristics; 2) establishing a binary hypothesis test model; 3) constructing a state value according to the dynamic physical layer fingerprint features, selecting and constructing a behavior value according to a threshold value, and establishing a state-behavior-benefit triple by taking a Bayesian risk function as an instantaneous benefit function; and 4) based on a depth deterministic strategy gradient framework, designing a detection threshold dynamic selection method, and detecting a physical layer spoofing attack. According to the method, dynamic continuous selection of the detection threshold can be realized, the method has adaptivity to a dynamic unknown environment, and the physical layer spoofing attack can be effectively detected.

The invention relates to the field of project risk decomposition and identification, and discloses a project risk decomposition identification method based on a safe operation basic function requirement, wherein the method settles a problem of high difficulty in performing accurate risk identification and logical relation carding. The method comprises the steps of performing system decomposition on a complicated project system from a high grade to a low grade so that mutual independence and completeness of a lower sub-system and an upper parent system are realized; finding out a to-be-satisfied function requirement of each sub-system based on the safe operation basic function requirement; analyzing possible events which affect realization of the function requirement; performing causal reasoning on each possible event, and finding out a reasoning event of the possible event; and analyzing the result of each reasoning event. The project risk decomposition identification method is suitable for complicated systematical projects such as water power engineering and civil engineering. The project risk decomposition identification method has an important application prospect in risk identification and analysis aspects of the complicated systematical projects and can realize a Bayesian risk network for constructing a complicated systematical project.

The invention discloses a homomorphic encryption method and application of the homomorphic encryption method to a privacy protection classifier. The method comprises the following specific steps: generating a public key pk and a private key sk by using a key generation algorithm KeyGen, encrypting a plaintext Q, and decrypting a ciphertext c. Compared with the prior art, the method has the advantages that the defect that multiplication can be carried out only when digits in a fixed-point representation system are represented as shared integers is overcome, a good balance effect is achieved between calculation efficiency and communication interaction, expected misclassification loss is reduced when the method is applied to privacy protection classifier encryption, and the privacy protection classifier encryption efficiency is improved. And the same accuracy rate as that of the original Bayesian classifier with the minimum Bayesian risk is kept.