63 results about "Online algorithm" patented technology

The present invention is directed to a mathematical approach to detect faults by reconciling known data driven techniques with a physical understanding of the HVAC system and providing a direct linkage between model parameters and physical system quantities to arrive at classification rules that are easy to interpret, calibrate and implement. The fault modes of interest are low system refrigerant charge and air filter plugging. System data from standard sensors is analyzed under no-fault and full-fault conditions. The data is screened to uncover patterns though which the faults of interest manifest in sensor data and the patterns are analyzed and combined with available physical system information to develop an underlying principle that links failures to measured sensor responses. These principles are then translated into online algorithms for failure detection.

The present invention is directed to a mathematical approach to detect faults by reconciling known data driven techniques with a physical understanding of the HVAC system and providing a direct linkage between model parameters and physical system quantities to arrive at classification rules that are easy to interpret, calibrate and implement. The fault modes of interest are low system refrigerant charge and air filter plugging. System data from standard sensors is analyzed under no-fault and full-fault conditions. The data is screened to uncover patterns though which the faults of interest manifest in sensor data and the patterns are analyzed and combined with available physical system information to develop an underlying principle that links failures to measured sensor responses. These principles are then translated into online algorithms for failure detection.

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.

An online algorithm exchange is presented herein. The online algorithm exchange provides access to a plurality of coloring algorithms from any number of developers. Users can select a particular algorithm for use as well as specify particular color measurements as inputs thereto. The color measurements can be selected from a database at the algorithm exchange, or provided by the user from a client terminal.

The invention belongs to the technical field of mobile edge computing unloading, and provides a task unloading method based on approximate optimization and reinforcement learning in MEC. A mathematical model is established for the problem, and a to-be-solved problem is converted into an integer linear programming problem. In order to solve the problem. an offline algorithm provided by the invention performs relaxation operation on integer constraint conditions in the problem, and then performs filtering and rounding operation on a solving result in sequence to obtain a final solution. The invention also provides an online algorithm for solving the problem, a linear regression method is used for predicting and providing an unloading strategy based on a reinforcement learning theory, and then a corresponding optimal resource allocation strategy is further provided by combining a deep neural network on the basis. A reasonable task unloading and resource allocation strategy can be formulated for a user under a limited resource condition, so that the application program execution delay and equipment energy consumption of user equipment are effectively reduced, and the utilization rateof the whole network is improved while the service quality is improved.

A distribution and scheduling system for advertisements that targets ads to users and maximizes service-provider revenue without having full knowledge of user-profile information. Each user device stores a user profile and is pre-loaded with a set of ads that could possibly be shown during a timeslot. Each user device selects and displays an ad based on the user profile but does not identify the selected ad to the service provider. Instead, the user devices provide perturbed user-profile information in the form of Boolean vectors, which the service provider uses in conjunction with a guaranteed-approximation online algorithm to estimate the number of users that saw a particular ad. Thus, the service provider can charge advertisers for the number of times their ads are viewed, without knowing the users' profiles or which ads were viewed by individual users, and users can view the targeted ads while maintaining privacy from the service provider.

The invention discloses a sky line online calculation method based on user clustering. The sky line online calculation method comprises the following steps of: (1) gathering users with similar QoS (Quality of Service) through user clustering, and substituting for personal characteristics of each user by using common characteristics of clustering to effectively reduce the data calculating volume, converting a mode of carrying out offline processing on all the users into a mode of carrying out offline processing on each cluster to improve the offline calculation efficiency, greatly reduce the storage space consumption and improve the enablement of a system; (2) dividing a calculation process into an offline module and an online module, carrying out a great quantity of complex calculations in the offline module to effectively improve the execution efficiency of an online algorithm; and (3) predicting an individual sky line by using a group characteristic sky line to obtain a prediction result, wherein on one hand, the prediction result is returned as a result when the requirement on precision is not high, and on the other handle, the prediction result can be corrected on the predicted sky line when an accurate sky line needs to be solved, and therefore the calculation process can be simplified.

The invention discloses a method for realtime online verification of a complex traffic control algorithm. According to the method, a traffic simulation software module, a database module and an algorithm implementation and evaluation module are provided, traffic state information can be acquired in real time and be provided to photo-realistic traffic control equipment under the condition of simulation, a traffic control scheme is generated by the photo-realistic traffic control equipment according to a verified algorithm and is applied to a simulation environment again, and realtime online verification is performed on the complex traffic control algorithm. Compared with the prior art, the method can be used for performing realtime online simulation verification on the complex traffic control algorithm, so guarantee is provided for the safe and effective application of the complex traffic control algorithm in reality; meanwhile, a method for online algorithm optimization is provided for algorithm researchers, and a practical and effective way is provided for future research work.

The invention provides a vehicle identification system and a vehicle identification method based on three-dimensional model and image matching. The method comprises the following steps of performing initial detection location to an appointed object in a real-time video by the online algorithm, and performing more accurate analysis and detection to the detection result by the offline algorithm. According to the invention, the three-dimensional model of the vehicle is converted to be a two-dimensional image with object structure information at a specific visual angle and then is subjected to matching detection with the image to be detected provided by the online algorithm; meanwhile, for the successful matching result, the structure information can be provided; by adoption of the vehicle identification method, the high accuracy of the online vehicle detection result; in addition, the disadvantages that the offline vehicle detection method is complex and can not be used in a real-time system are overcome; and the purpose of obtaining real-time accurate detection results can be achieved by combining the online and offline detection methods to form complementary of algorithms.

The invention provides a fetal electrocardiosignal self-adaptive blind extraction method based on generalized eigenvalue maximization. The method comprises the following steps: collecting and acquiring blended data including a maternal and fetal electrophysiological signal; based on the self-adaptive real-time algorithm, calculating the covariance matrix characteristic value and characteristic vector of different delayed autocorrelation matrixes within the set period range; selecting the characteristic vector corresponding to the maximum eigenvalue as the blind separation vector so as to extract a fetal electrocardiosignal. The fetal electrocardiosignal self-adaptive blind extraction method utilizes the second-order statistics property of the signal to minimize the noise-to-signal ratio. The second-order statistic method is simple in calculation and small in operation, can still realize effective calculation when the rank of the autocorrelation matrix of the observation vector is close to zero, and can expand to real-time on-line algorithm more conveniently to facilitate product development.

The invention belongs to the field of pattern recognition and computer vision and discloses a target tracking method with temporal and spatial context information combined. The method comprises steps: a first frame of picture is used to train an initial strong classifier, and a temporal and spatial context model needed by next-frame tracking is learnt; when a new frame arrives, the well-trained strong classifier is used to estimate a plurality of blocks in a search area, and a first confidence matrix is obtained; the temporal and spatial context information is then integrated to obtain a confidence graph function, the function is used to solve the confidence value of each block in the search area, and a second confidence matrix is obtained; and finally, according to corresponding weight linear combination, a final confidence matrix is obtained, and a block with the maximum confidence value in the final confidence matrix is found to be the tracked target. Through combining the temporal and spatial context information of the target to an online Boosting algorithm, quick robust tracking can be realized.

The invention relates to a small group real-time detection method of an intermediate-density scene. The method includes: inputting a video and individual movement track detection data; obtaining a detection coordinate position and a motion speed of each individual in each frame according to the individual movement track data; establishing a target prediction model for each motion individual; calculating a target direction [eta]i of each motion individual; and substituting [eta]i into a correlation filtering algorithm for small group detection to obtain a correlation small group. The method is applicable to various types of intermediate-density dense scenes, the operation speed is fast, real-time operation can be realized, the accuracy and the recall rate in various scenes are high, and the method is an online algorithm so that offline learning is avoided.

The invention discloses a dynamic vehicle scheduling and route planning method for a shared bus. The method comprises the following steps: abstracting the operation cost of an actual bus company and the travel experience of passengers, and establishing a multi-objective optimization model; extracting a traffic topology suitable for the operation of a shared bus according to a real application areaof the shared bus and the road network; generating a candidate line set of the shared bus by using an improved local search algorithm based on the extracted traffic topology; designing an offline algorithm to solve the problems of shared bus scheduling and route planning in application scenarios with similar passenger flow rules; designing an online algorithm to solve the problems of shared bus scheduling and route planning of application scenarios with dynamic and real-time passenger flow. The invention provides a new method for dynamic scheduling and route planning of the shared bus, the contradiction between the operation cost and the passenger travel experience faced by a shared bus company is truly solved, and a new effective scheme is provided for vehicle scheduling and route planning of the shared bus company.

The invention discloses a hierarchical reinforcement learning method and device based on strategy options. The method comprises the following steps: constructing a high-level strategy network, a low-level strategy network and an evaluation network; acquiring a state track from the simulation environment; updating the parameters of the high-level strategy network, the low-level strategy network andthe evaluation network based on the state trajectory and the learning process of the strategy online algorithm; generating an updated strategy model according to the high-level strategy network, thelow-level strategy network and the evaluation network of which the parameters are updated, and testing the updated strategy model. According to the hierarchical reinforcement learning method based onthe strategy option, the action and the high-level strategy can be learned from zero in the simulation environment, the performance is stable, and the data use efficiency is high.

The invention discloses an energy-consumption-optimization adaptive streaming media distribution method suitable for a 4G environment. The method comprises the following steps: designing an adaptive streaming media distributor suitable for the 4G environment, performing a deployment on a gateway, combining data blocks into burst streams, issuing the burst streams to users, and optimizing network throughput, watching experiences of the users and energy consumption of user equipment in a combined way; designing a utility function for calculating user utilities, including a video definition and penalties of user experiences, of unit energy consumption, in order to realize proportional fairness among a plurality of users and smooth playing; modeling an optimal user utility problem into a non-linear integer programming problem and giving an efficient online allocation algorithm; executing the online algorithm based on channel quality and a caching state of the user equipment in order to allocate resources to the users; and affecting bit rate selections of the users through the resource allocation in order to finally realize effective video transmission under a multi-user situation.

The invention relates to a parking space allocation method based on a distribution technology. The method comprises steps of constructing a database, collecting user demands, preprocessing data, operating an online algorithm, sending an allocation result and navigating a terminal. In the step of constructing a database, user data is stored by using table structures based on Mysql and HBase databases and databases thereof; in the step of collecting user demands, parking requests of users in a normalized timeslot are collected, and the user uses an intelligent terminal to select a destination through searching and clicking manners; and in the step of preprocessing data, in the stage at when a server receives user demands, all data information about user demands is collected in the normalizedtimeslot and users and related parking spaces are numbered. The parking space allocation method has the beneficial effects that an available parking space close to the destination can be provided fora car owner, the time for searching the parking space by the car owner is reduced greatly, the cost for searching the parking space is reduced, and the method has benefits in optimization of urban traffic environment and parking benefits.

A distribution and scheduling system for advertisements that targets ads to users and maximizes service-provider revenue without having full knowledge of user-profile information. Each user device stores a user profile and is pre-loaded with a set of ads that could possibly be shown during a timeslot. Each user device selects and displays an ad based on the user profile but does not identify the selected ad to the service provider. Instead, the user devices provide perturbed user-profile information in the form of Boolean vectors, which the service provider uses in conjunction with a guaranteed-approximation online algorithm to estimate the number of users that saw a particular ad. Thus, the service provider can charge advertisers for the number of times their ads are viewed, without knowing the users' profiles or which ads were viewed by individual users, and users can view the targeted ads while maintaining privacy from the service provider.

The invention discloses a method and device for managing an algorithm model, and relates to the technical field of computers. A specific embodiment of the method comprises the steps of receiving configuration information for an algorithm model management container, wherein the algorithm model management container is a software platform used for placing an algorithm model, and the configuration information comprises an identifier of the required algorithm model; obtaining an identifier of a loaded algorithm model in the algorithm model management container; and updating the algorithm model management container based on the identifier of the required algorithm model and the identifier of the loaded algorithm model. According to the implementation mode, the algorithm service version needing to be called can be dynamically loaded, and reconfiguration and deployment are not needed when the version is switched, and meanwhile restarting is not needed when the algorithm model is loaded, unloaded or updated; and the memory consumption of the server is reduced, and the bandwidth is saved, and the performance of online algorithm model service is improved.

The invention is applicable to the technical field of positioning and provides a debugging method and device for online positioning and a positioning system. The debugging method comprises the steps that fault reporting information during online positioning is received, wherein the fault reporting information includes corresponding online problems; whether corresponding basic satellite data goes wrong or not is judged; when it is determined that the basic satellite data does not go wrong, an algorithm log is obtained, wherein the algorithm log includes at least online algorithm log; based on the basic satellite data and the algorithm log, problem prompts are given. According to the debugging method, when online positioning goes wrong, whether the basic satellite data goes wrong or not is analyzed firstly, problem causes are analyzed according to the algorithm log when the basic satellite data does not go wrong, multiple factors of the positioning process are considered, and the problemchecking efficiency is improved.

The invention relates to an online cross-scale multi-fluid target matching and tracking method, which comprises the following steps of: extracting an associated object by adopting a sliding window, matching and tracking specific characteristics on an associated object image, and extracting fluid target characteristics of different scales; selecting basic characteristic parameters for characteristic extraction according to fluid target characteristics of different scales; calculating a composite association weight according to the feature information of the two continuous frames, extracting spatial association and time association information between different body targets through the composite association weight, and constructing a weighted bipartite graph based on the two adjacent framesbased on the association information; performing online target matching by adopting a sparse weighted bipartite graph matching algorithm, and constructing a sparse weighted bipartite graph based on the matched target; and performing target tracking and trajectory prediction by adopting a Kalman filtering algorithm based on the constructed sparse weighted bipartite graph. According to the invention, the multi-fluid target matching tracking precision can be effectively improved; the running time of the method is reduced to the second level, and the online algorithm requirement is met.

The invention relates to a wireless context aware network, in particular to a context-aware information centralization resource management method in a multi-hop cellular network architecture. A new structure is provided in information centralization and the multi-hop cellular network. Communication traffic is reduced in the network by introducing caches in the network and processes in the route. Considering priority, delay and size of information packets and Rayleigh attenuation in a wireless link, an optimal resource management method is provided to minimize long-term energy consumption. The optimization problem is modeled into a limited restless bandits problem which is further modeled into a standard restless bandits model, and complexity in online algorithms can be reduced with the aid of retrievable characteristic of the model. By the method, performance of the context aware network is improved significantly.

The invention provides a test method and device, after a network object and network object network behavior data are set as input data, the input data are processed by an on-line algorithm routine, the data result is obtained, further the value of a target field is extracted from the data result, sample data are obtained, then the sample data are evaluated according to the test index, the test result is obtained, only the target field is required to be set according to the existing test index, and the input data are set, so that the test index can be used for testing the on-line algorithm routine, the process of editing a simulation script in the prior art is not required, and accordingly, the test efficiency of the on-line algorithm routine is improved.

The invention relates to an idling start and stop system based on online algorithm optimization. The system comprises a central processing module (1), a power supply module (2), a data acquiring module (3), a communication module (4) and a driving module (5), and is characterized in that the central processing module (1) is responsible for acquiring signals of vehicle state parameters, exchanges data with the data acquiring module (3), the communication module (4) and the driving module (5) through a data bus, and analyzes switch signals and simulation signals of the vehicle state parameters based on an online algorithm to determine if a vehicle engine is located in a closing state or an opening state so as to send an instruction of controlling opening or closing of the engine to a ESM starter module and an electromagnetic valve driving circuit module. Before the idling start and stop system is operated, the idling start and stop are prejudged, so that the operation frequency of the system is reduced, and the control confusion of the idling start and stop system is reduced.

The disclosure relates to a method and system for intelligent visualization in the monitoring of process and / or system data used in technical processes and in the operation of technical systems. The behavior of signals in the past is analyzed and used for a future optimized visualization. Continuously running online algorithms support the system operator by detecting and correspondingly highlighting deviations from the historically observed signal patterns.

The invention discloses an online soft measurement method for a sand return ratio in an ore grinding classification process, which comprises the following steps: firstly, detecting the ore feeding flow, pressure and concentration of a cyclone in real time, and filing real-time detection data into a database; then, completing soft measurement algorithm parameter calculation; selecting a soft measurement calculation period of the sand return ratio; and finally, calculating the sand return ratio on line. The online algorithm only depends on online measurement values of ore feeding flow, pressureand concentration of the cyclone, and online soft measurement of the sand return ratio is achieved; parameters are determined by using an off-line multi-sampling method, and a least square method is used during parameter calculation, so that the minimization of a sand return ratio calculation error is ensured.

In the invention, how to combine energy collection with a large-scale multi-antenna system is studied, so that the energy efficiency is obviously improved, and meanwhile, the requirements of actual limiting causal relationships such as energy and the like, RF chain energy consumption, limited battery capacity, transmission power and QoS (Quality of Service) are met. From a deterministic offline hypothesis, a field of view with limited united antenna selection and power distribution problems is formulated, which is due to the combined NP difficult property. We propose a nonlinear offline algorithm-combined fractional programming, dichotomy and Lagrangian dual decomposition, which is the performance reference. Then, comprehensive property analysis is provided, and an offline algorithm is provided from the aspects of optimality, convergence, uniqueness and complexity. By exploring a solution structure of an offline algorithm, we propose low-complexity causal knowledge of a suboptimal online algorithm which is only needed. And finally, the proposed offline and online algorithms pass simulation verification and comparison of various system configurations.

The embodiment of the invention provides an in-cloud collaborative data sharing method and device for mobile equipment, and the method comprises the steps: a data transmission scheduling problem in the cloud of the mobile equipment being formalized into an efficiency maximization problem of data transmission in the cloud of the mobile equipment based on a cloud data sharing model of the mobile equipment; wherein in the efficiency maximization problem, a cooperation excitation mechanism, service quality requirement QoE heterogeneity of different mobile nodes and a multi-interface communicationmode are used as constraint conditions; wherein the cooperation excitation mechanism is used for exciting the mobile nodes to participate in data transmission cooperation, and QoE heterogeneity of different mobile nodes indicates that different mobile nodes have different QoE requirements; wherein the mobile node is the mobile device; and solving the efficiency maximization problem through an online algorithm to obtain an optimal auxiliary variable decision, an optimal data transmission decision and an optimal interface selection decision, so that the data transmission efficiency in the mobileequipment cloud can be maximized.

A fast division method which uses a smaller quotient digit set of {−1, 1} than {−1, 0, 1} that is used by known algorithms, therefore accelerates the speed of calculation. Partial remainders can be computed with the signals of remainders decided independently and in parallel. By taking the absolute values of the remainders, we can successively subtract the remainders without the need of knowing the signs of remainders, while signs of the remainders can be decided in parallel and independently at the same time. The algorithm adopts non-restoring division operation and CSA type of operation for fast subtraction. The algorithm is also an on-line algorithm that facilitates highly pipelined operation while it is much simpler than the existing on-line algorithms.

The invention discloses a manufacturing service resource dynamic scheduling method and system based on a random online algorithm, and the method and system are applied to a server. The method comprises: storing the numbers of all manufacturer clients, the distances between which and purchaser clients are within a set range, in a container; receiving a production target and production resources submitted by a purchaser client connected with the server; storing the received production target and the production resource in a resource pool; receiving production state data submitted by a plurality of manufacturer clients connected with the server; traversing the production state data submitted by all manufacturer clients; storing the production state data of each manufacturer at each moment in a queue; and according to the production state data of each manufacturer at each moment, randomly selecting a manufacturer capable of realizing a production target, and pushing production resources in a production cycle to a corresponding manufacturer client.

The invention discloses a guarantee method for a basic data quality algorithm of a network platform. The method comprises the steps of by utilizing sample data, extracting multiple pieces of tagged data from basic data to serve as the sample data; performing offline algorithm regression assessment on a tested algorithm by utilizing the sample data; and through the sample data accumulated in the basic data, performing online algorithm regression assessment on the tested algorithm. While the tested algorithm is subjected to online and offline testing, a feedback is generated for the algorithm, the iteration of the algorithm is performed, and the algorithm continues to be synchronously verified after correction is finished; and the effect of the algorithm is subjected to reliable measurement,and the effect is taken for the iteration of the algorithm, so that the quality of the basic data is ensured.