13905results about How to "Fast convergence" patented technology

Reduction of administrative overhead in maintaining network information, rapid convergence on an optimal routing path through the data network, and utilization of only required network resources are realized by a novel method for establishing a call path between network users. The method is based upon deployment of a network information server that stores network topology information and that is addressable by each end user. In this method, the network information server receives a request to establish a call path. The request identifies at least the calling party. In response to the request, the network information server determines a network traversal between the calling party and a root network wherein the network traversal includes call path information about the sub-networks between the calling party and the root network. The request for establishing a call path can also identify the called party. Based on the calling and called party identification, the network information server also determines a second network traversal between the called party and the root network. The second network traversal is sent to either the calling party or the called party or to both the calling and called parties. The server can determine an intersection of the traversals and send the intersection information to the parties. The intersection information is known as a merge point and represents an optimal call path between the parties.

A serverless name resolution protocol ensures convergence despite the size of the network, without requiring an ever-increasing cache and with a reasonable numbers of hops. This convergence is ensured through a multi-level cache and a proactive cache initialization strategy. The multi-level cache is built based on a circular number space. Each level contains information from different levels of slivers of the circular space. A mechanism is included to add a level to the multi-level cache when the node determines that the last level is full. A peer-to-peer name resolution protocol (PNRP) includes a mechanism to allow resolution of names which are mapped onto the circular number space through a hash function. Further, the PNRP may also operate with the domain name system by providing each node with an identification consisting of a domain name service (DNS) component and a unique number.

A serverless name resolution protocol ensures convergence despite the size of the network, without requiring an ever-increasing cache and with a reasonable numbers of hops. This convergence is ensured through a multi-level cache and a proactive cache initialization strategy. The multi-level cache is built based on a circular number space. Each level contains information from different levels of slivers of the circular space. A mechanism is included to add a level to the multi-level cache when the node determines that the last level is full. A peer-to-peer name resolution protocol (PNRP) includes a mechanism to allow resolution of names which are mapped onto the circular number space through a hash function. Further, the PNRP may also operate with the domain name system by providing each node with an identification consisting of a domain name service (DNS) component and a unique number.

A serverless name resolution protocol ensures convergence despite the size of the network, without requiring an ever-increasing cache and with a reasonable numbers of hops. This convergence is ensured through a multi-level cache and a proactive cache initialization strategy. The multi-level cache is built based on a circular number space. Each level contains information from different levels of slivers of the circular space. A mechanism is included to add a level to the multi-level cache when the node determines that the last level is full. A peer-to-peer name resolution protocol (PNRP) includes a mechanism to allow resolution of names which are mapped onto the circular number space through a hash function. Further, the PNRP may also operate with the domain name system by providing each node with an identification consisting of a domain name service (DNS) component and a unique number.

A serverless name resolution protocol ensures convergence despite the size of the network, without requiring an ever-increasing cache and with a reasonable numbers of hops. This convergence is ensured through a multi-level cache and a proactive cache initialization strategy. The multi-level cache is built based on a circular number space. Each level contains information from different levels of slivers of the circular space. A mechanism is included to add a level to the multi-level cache when the node determines that the last level is full. A peer-to-peer name resolution protocol (PNRP) includes a mechanism to allow resolution of names which are mapped onto the circular number space through a hash function. Further, the PNRP may also operate with the domain name system by providing each node with an identification consisting of a domain name service (DNS) component and a unique number.

The invention discloses a fundus image retinal vessel segmentation method and a fundus image retinal vessel segmentation system based on deep learning. The fundus image retinal vessel segmentation method comprises the steps of performing data amplification on a training set, enhancing an image, training a convolutional neural network by using the training set, segmenting the image by using a convolutional neural network segmentation model to obtain a segmentation result, training a random forest classifier by using features of the convolutional neural network, extracting a last layer of convolutional layer output from the convolutional neural network, using the convolutional layer output as input of the random forest classifier for pixel classification to obtain another segmentation result, and fusing the two segmentation results to obtain a final segmentation image. Compared with the traditional vessel segmentation method, the fundus image retinal vessel segmentation method uses the deep convolutional neural network for feature extraction, the extracted features are more sufficient, and the segmentation precision and efficiency are higher.

A conventional hill climbing method is not capable of tracking a maximum power when a shadow covers a part of a solar cell panel and a plurality of maximal values is thereby developed. The invention provides a photovoltaic power generation system control method that controls a photovoltaic output setting value to be an appropriate value by following up a variation in the solar irradiation, including a genetic algorithm process of employing the photovoltaic output setting value as a gene and the output power as an evaluation value thereof; randomly extracting a plurality of genes and repeating storage, selection and crossover / mutation of the output powers which are the evaluation values of the genes, so as to converge a difference between a maximum and minimum values of the output voltages into a predetermined range; and a hill climbing process of setting the photovoltaic output setting value corresponding to the greatest value of the converged output voltages, and a greater and a lower value than the photovoltaic output setting value by a predetermined amount, so as to select the greatest output power among these values; and repeating such steps to track the maximum output power.

The invention provides a hardware neural network conversion method which converts a neural network application into a hardware neural network meeting the hardware constraint condition, a computing device, a compiling method and a neural network software and hardware collaboration system. The method comprises the steps that a neural network connection diagram corresponding to the neural network application is acquired; the neural network connection diagram is split into neural network basic units; each neural network basic unit is converted into a network which has the equivalent function with the neural network basic unit and is formed by connection of basic module virtual bodies of neural network hardware; and the obtained basic unit hardware networks are connected according to the splitting sequence so as to generate the parameter file of the hardware neural network. A brand-new neural network and quasi-brain computation software and hardware system is provided, and an intermediate compiling layer is additionally arranged between the neural network application and a neural network chip so that the problem of adaptation between the neural network application and the neural network application chip can be solved, and development of the application and the chip can also be decoupled.

Adaptive antenna array techniques for use in an orthogonal frequency division multiplexed spread-spectrum multi-access (OFDM-SSMA) cellular wireless system or other type of wireless communication system. A base station of the system includes an antenna array and a base station receiver. The base station receiver implements an adaptive antenna gain algorithm which estimates a spatial covariance matrix for each of K mobile stations communicating with the base station. The spatial covariance matrix for a given one of the mobile stations is determined at least in part based on a unique hopping sequence of the mobile station, and provides a correlation between signals received from the mobile station at different antenna elements within the antenna array. An average spatial covariance matrix for a set of received signals is also generated. The individual spatial covariance matrices and the average spatial covariance matrix are processed to generate an estimate of an interference matrix for the K mobile stations, and the estimate of the interference matrix is further processed to generate array responses for each of the mobile stations. The array response for a given mobile station is processed to determine an antenna weighting which is applied to a signal received from the given mobile station in order to facilitate detection of a corresponding transmitted symbol.

The invention provides a pose estimation method based on RGB-D and IMU information fusion. The method comprises the following steps: S1, after time synchronization of RGB-D camera data and IMU data, the gray image and depth image acquired by the RGB-D camera and the acceleration and angular velocity information collected by the IMU are preprocessed to obtain the characteristics of adjacent frame matching in the world coordinate system. Point and IMU state increment; S2, a visual inertial device in a system is initialized according to system external parameters of a pose estimation system; S3,according to information of the intilized visual inertial device, feature points matching adjacent frames in a global corrdinate system, least squares optimization functions of an IMU state incrementconstruction system; an optimization method is used to iteratively solve the optimal solution of the least squares optimization function, and the optimal solution is used as the pose estimation statequantity; further, loop detection is performed to acquire globally-consistent pose estimation state quantity. Therefore, feature point depth estimation is more accurate, and positioning precision of the system is improved.

A system and method for identifying minor echoes present in an input signal in the situation where a set of major echoes has already been identified from the input signal. The method includes: computing a spectrum F corresponding to a sum of the major echoes; computing a weighted power spectrum SM of the spectrum F; subtracting the weighted power spectrum SM from a weighted power spectrum PIN of the input signal to obtain a difference spectrum; performing a stabilized division of the difference spectrum by a conjugate of the spectrum F to obtain an intermediate spectrum; computing an inverse transform of the intermediate spectrum to obtain a time-domain signal; and estimating parameters one or more of the minor echoes from the time-domain signal. The echo parameters are usable to remove at least of portion of the one or more estimated minor echoes from the input signal.

Some embodiments provide a method for providing redundancy and fast convergence for modules operating in a network. The method configures modules to use a same anycast inner IP address, anycast MAC address, and to associate with a same anycast VTEP IP address. In some embodiments, the modules are operating in an active-active mode and all nodes running modules advertise the anycast VTEP IP addresses with equal local preference. In some embodiments, modules are operating in active-standby mode and the node running the active module advertises the anycast VTEP IP address with higher local preference.

Significant, cost-effective improvement is introduced for Position, Navigation, and Timing (PNT) on a global basis, particularly enhancing the performance of Global Navigation Satellite Systems (GNSS), an example of which is the Global Positioning System (GPS). The solution significantly improves performance metrics including the accuracy, integrity, time to acquire, interference rejection, and spoofing protection. A constellation of small satellites employing a low-cost architecture combined with improved signal processing yields an affordable enabler for spectrum-efficient transportation mobility. As air traffic management modernization transitions to a greater dependence on satellite positioning, the solution provides aviation users new protections from both intentional and unintentional interference to navigation and surveillance. And in response to an era in which intelligent transportation is under development for automobiles, reliable where-in-lane positioning enables new applications in connected and autonomous vehicles. New military capability increases PNT availability.

The invention provides an autonomous underwater vehicle (AUV) trajectory tracking control method based on deep reinforcement learning, belonging to the field of deep reinforcement learning and intelligent control. The autonomous underwater vehicle trajectory tracking control method based on deep reinforcement learning includes the steps: defining an AUV trajectory tracking control problem; establishing a Markov decision-making process model of the AUV trajectory tracking problem; constructing a hybrid policy-evaluation network which consists of multiple policy networks and evaluation networks;and finally, solving the target policy of AUV trajectory tracking control by the constructed hybrid policy-evaluation network, for the multiple evaluation networks, evaluating the performance of eachevaluation network by defining an expected Bellman absolute error and updating only one evaluation network with the lowest performance at each time step, and for the multiple policy networks, randomly selecting one policy network at each time step and using a deterministic policy gradient to update, so that the finally learned policy is the mean value of all the policy networks. The autonomous underwater vehicle trajectory tracking control method based on deep reinforcement learning is not easy to be influenced by the bad AUV historical tracking trajectory, and has high precision.

A technique configures an intermediate network node to automatically determine whether a route advertised by a routing protocol is important for fast convergence in a computer network. As used herein, an important route needed for fast convergence is a route advertised by the routing protocol, such as an exterior gateway routing protocol (EGP) process executing on the node, as a next-hop address, since external connectivity relies on such a route. Notably, the EGP process interacts with an interior gateway routing protocol (IGP) process executing on the node to identify the route as an important route. Identification of an important route, in turn, allows IGP to process the route in a high priority fashion, thereby facilitating fast convergence.

A method for fast converging an end-to-end service and a Provider Edge (PE) includes: setting routing information of at least two tunnels in a double-ascription PE of a remote Customer Edge (CE), wherein, the two tunnels are from the double-ascription PE of the remote CE to the PE connected with the remote CE; detecting tunnel states to obtain state information of the tunnels; the double-ascription PE obtaining available routing information and routing information of the at least two tunnels, and forwarding the service according to the available routing information. The double-ascription PE of the remote CE can directly forward the service according to the pre-configured routing information of other tunnels when the current tunnel is unavailable, such as a terminal node of the current tunnel is abnormal, thereby avoids the procedure of re-selecting the route, and increases the end-to-end service convergence speed and improves the service reliability.

Method and apparatus for performing close-loop programming of resistive memory devices in crossbar array based hardware circuits and systems. Invention provides iterative training of memristor crossbar arrays for neural networks by applying voltages corresponding to selected training patterns. Error is detected and measured as a function of the actual response to the training patterns versus the expected response to the training pattern.

A targeted advertising system uses a machine learning tool to select an asset for a current user of a user equipment device, for example, to select an ad for delivery to a current user of a digital set top box in a cable network. The machine learning tool first operates in a learning mode to receive user inputs and develop evidence that can characterize multiple users of the user equipment device audience. In a working mode, the machine learning tool processes current user inputs to match a current user to one of the identified users of that user equipment device audience. Fuzzy logic may be used to improve development of the user characterizations, as well as matching of the current user to those developed characterizations. In this manner, targeting of assets can be implemented not only based on characteristics of a household but based on a current user within that household.

A convergence prism assembly with a housing and a prism structure having a plurality of prisms mounted in a unitary assembly.

Methods and devices for determining optimal Atrial to Ventricular (AV) pacing intervals and Ventricular to Ventricular (VV) delay intervals in order to optimize cardiac output. Impedance, preferably sub-threshold impedance, is measured across the heart at selected cardiac cycle times as a measure of chamber expansion or contraction. One embodiment measures impedance over a long AV interval to obtain the minimum impedance, indicative of maximum ventricular expansion, in order to set the AV interval. Another embodiment measures impedance change over a cycle and varies the AV pace interval in a binary search to converge on the AV interval causing maximum impedance change indicative of maximum ventricular output. Another method varies the right ventricle to left ventricle (VV) interval to converge on an impedance maximum indicative of minimum cardiac volume at end systole. Another embodiment varies the VV interval to maximize impedance change.

As no single wireless network can provide both broadband high-speed radio transmission and seamless full mobility for commercial wireless communications in terms of technology, cost, spectrum and performance, a new wireless converged network access infrastructure based on Open Wireless Architecture (OWA) has been disclosed, wherein the spectrum is separated into low spectrum band (LSB) for wide area seamless mobility networks and high spectrum band (HSB) for very high-speed broadband wireless access systems in order to maximize the spectrum utilization efficiency for the converged multiple wireless access networks of the fully service-oriented wireless communications including multimedia broadcast and multicast service.

There is provided a head mounted display including a glasses-type frame to be worn on a head of an observer; two optical modules including two image creation devices, and two light guides having two light guide plates coupled one-to-one with the two image creation devices and placed closer to a center of a face of the observer than the image creation devices are as a whole, that guide light beams output from the image creation devices and output the light beams toward pupils of the observer; and an optical plate supporting the two light guides, wherein the optical plate is attached to a center part of the frame.

The invention relates to a multi-factory cooperative scheduling optimization method during equipment manufacturing. The method is a multi-agent and improved ant colony algorithm-based scheduling model and algorithm and is characterized by comprising the following steps of: establishing a multi-agent-based scheduling system model framework; introducing a genetic algorithm into the ant colony optimization process and establishing the improved ant colony algorithm to support the scheduling decisions of the agents and determine production equipment, a processing order and processing time for each workpiece task of a manufacturing system. The method has the advantages of optimizing the target, shortening the completion time of the workpiece task, improving the utilization rate and production efficiency of the equipment along with simple operation, high convergence rate and high convergence performance; moreover, the method is suitable for a multi-factory cooperative manufacturing process and the common production process of the common manufacturing enterprise as well, realizes coordination among a plurality of manufacturers producing the same product during the manufacturing by optimal task allocation and scheduling, has a wide application range and is suitable for popularization and application.

An endoscopic system and method that is adaptable for therapeutic applications as well as sensor operation and is capable of producing 3-dimensional human vision simulated imaging with real dynamic convergence, not virtual convergence. Applications may include use in any space, including but not limited to, intra-abdominal cavities, intra-thoracic cavities, and intra-cranial cavities. Further, two or more diagnostic / sensor probes may be used, with at least two being the same kind to create the 3-dimensional effect, such as but not limited to, camera, ultrasound, and magnetic-resonance imaging. Diagnostic / sensor probes are each mounted to the end of a different arm, with the other ends of the two arms both being attached to the same hinge that allows them to turn freely on the same axis from side-to-side within a 180 degree angle range of movement on the distal end of a main tubular shaft system. Medical, as well as other applications, are contemplated.

Methods and apparatus which characterize the ionospheric error across a network of GNSS reference stations are presented. The method relies on dual-frequency phase measurements in a geometry-free linear combination. The data are filtered for ambiguities and the characteristic parameters of the ionosphere. In combination with filter results from other combinations of phase measurements (ionosphere free combination), the physically-based model provides rapid and reliable ambiguity resolution.

The invention discloses a cascaded residual error neural network-based image denoising method. The method comprises the following steps of building a cascaded residual error neural network model, wherein the cascaded residual error neural network model is formed by connecting a plurality of residual error units in series, and each residual error unit comprises a plurality of convolutional layers, active layers after the convolutional layers and unit jump connection units; selecting a training set, and setting training parameters of the cascaded residual error neural network model; training the cascaded residual error neural network model by taking a minimized loss function as a target according to the cascaded residual error neural network model and the training parameters of the cascaded residual error neural network model to form an image denoising neural network model; and inputting a to-be-processed image to the image denoising neural network model, and outputting a denoised image. According to the cascaded residual error neural network-based image denoising method disclosed by the invention, the learning ability of the neural network is greatly enhanced, accurate mapping from noisy images to clean images is established, and real-time denoising can be realized.

An extended depth of field is achieved by a computational imaging system that combines a multifocal imaging subsystem for producing a purposefully blurred intermediate image with a digital processing subsystem for producing a recovered image having an extended depth of field. The multifocal imaging system preferably exhibits spherical aberration as the dominant feature of the purposeful blur. A central obscuration of the multifocal imaging subsystem renders point-spread functions of object points more uniform over a range of object distances. An iterative digital deconvolution algorithm for converting the intermediate image into the recovered image contains a metric parameter that speeds convergence, avoids stagnations, and enhances image quality.

Input measurements from a measurement device are processed as a Markov chain whose transitions depend upon the signal. The desired information related to the device can then be obtained by estimating the state of the signal at a time of interest. A nonlinear filter system can be used to provide an estimate of the signal based on the observation model. The nonlinear filter system may involve a nonlinear filter model and an approximation filter for approximating an optimal nonlinear filter solution. The approximation filter may be a particle filter or a discrete state filter for enabling substantially real-time estimates of the signal based on the observation model. In one applications a click stream entered with respect to a digital set top box of a cable television network is analyzed to determine information regarding users of the digital set top box so that ads can be targeted to the users.

The embodiment of the invention provides a neural network training and construction method and device for object detection, and an object detection method and device based on a neural network. The neural network training and construction method for object detection comprises the steps: inputting a training image comprising a training object into the neural network, so as to obtain a prediction boundary frame of the training object; obtaining a first loss function according to the ratio of the intersection area and union area of the prediction boundary frame and a real boundary frame, wherein the real boundary frame is the boundary frame of the training object marked in the training image in advance; and adjusting the parameters of the neural network at least through the first loss function, so as to carry out the training of the neural network. According to the invention, the first loss function is employed for the regression of a target boundary frame into an integrated unit, and the object detection precision of the neural network is remarkably improved. Moreover, the training and detection efficiency of the neural network can be effectively improved through two branch structures of the neural network.

The invention relates to an intelligent traffic control system and method based on a wireless Mesh ad hoc network. The control system comprises n traffic control nodes, m ZigBee area base stations, a mobile communication base station and a traffic control center, wherein the n traffic control nodes are positioned in all intersections, and all kinds of equipment in the n traffic control nodes are connected by an RS-232 / RS-485 bus; adjacent nodes realize a wireless Mesh ad hoc communication network by a ZigBee technology and exchange traffic data in real time; remote connection is carried out between each ZigBee area base station and the traffic control center by adopting GPRS / CDMA (General Packet Radio Service / Code Division Multiple Access) data transmission equipment through the mobile communication base station; and m and n are respectively integers of 1-100 and 2-256. The system organically combines an advanced ZigBee short-distance wireless communication technology and a modern intelligent network technology; and the control method of the invention adopts a self-coordinating intelligent distribution type control algorithm and can realize the optimal traffic signal control effect.