15722 results about "Network structure" patented technology

The invention relates to a cascaded depth neural network-based face attribute recognition method. The method includes the following steps that: 1) a cascaded depth neural network composed of a plurality of independent convolution depth neural networks is constructed; 2) a large number of face image data are adopted to train networks at all levels in the cascaded depth neural network level by level, and the output of networks of previous levels is adopted as the input of networks of posterior levels, such that a coarse-to-fine neural network structure can be obtained; and 3) the coarse-to-fine neural network structure is adopted to recognize the attributes of an inputted face image, and final recognition results can be outputted. According to the cascaded depth neural network-based face attribute recognition method of the invention, a cascade algorithm system is adopted based on depth learning, and therefore, training time can be accelerated; and a cascaded coarse-to-fine processing process is realized, and the performance of a final network can be improved by networks of each level through utilizing information of networks of upper levels, and therefore, the speed and the accuracy of face attribute recognition can be effectively improved.

A system and method for replacing file system processors, also known as hot swapping, is described. The system and method operate on a fault-tolerant network file system that includes a first file server that is operably connected to a network fabric and a second file server that is operably connected to the network fabric. The fault-tolerant network file system includes a first disk array that is operably coupled to the first file server and to the second file server and a second disk array that is operably coupled to the first file server and to the second file server. First file system information is loaded on the first file server. The first file system information includes a first intent log of proposed changes to first metadata. Second file system information is loaded on the second file server. The second file system information includes a second intent log of proposed changes to second metadata. The first file server has a copy of the second metadata, and the second file server maintains a copy of the first metadata, thereby allowing the first file server to access files on the second disk array in the event of a failure of the second file server.

The present invention combines a data processing structure with a graphical user interface (GUI) to create an information analysis tool wherein multiple functions are combined in a network to extract information from multiple data sources. The functional network is created, and graphically represented to the user, by linking individual operations together. The combination of individual operations is not limited by the input or output characteristic of any single operation. The form of the input to or output from a by individual operation, whether from a database or from another operation, is the same. That is, both the input to and the output from an analysis function is a list of document identifiers and corresponding document characteristics. Because the form of the input and output from each operation is the same, arbitrary combinations into of operations may be created. Moreover, functional networks of individual operations can then be used for database retrieval as well as to filter data streams. Furthermore, the user is able to create a visual representation of the structure forming a functional network which may be dynamically updated as new data is added or functions switched in or out. Because, inter alia, the network structure dynamically responds to information as it is presented to the network, the visual representation of the network conveniently provides the user with information concerning the characteristics of the database or stream of data that are substantially unavailable through conventional search, filtering, or clustering techniques alone.

The present invention provides a compressed low-resolution image restoration method based on a combined deep network, belonging to the digital image / video signal processing field. The compressed low-resolution image restoration method based on the combined deep network starts from the aspect of the coprocessing of the compression artifact and downsampling factors to complete the restoration of a degraded image with the random combination of the compression artifact and the low resolution; the network provided by the invention comprises 28 convolution layers to establish a leptosomatic network structure, according to the idea of transfer learning, a model trained in advance employs a fine tuning mode to complete the training convergence of a greatly deep network so as to solve the problems of vanishing gradients and gradient explosion; the compressed low-resolution image restoration method completes the setting of the network model parameters through feature visualization, and the relation of the end-to-end learning degeneration feature and the ideal features omits the preprocessing and postprocessing; and finally, three important fusions are completed, namely the fusion of the feature figures with the same size, the fusion of residual images and the fusion of the high-frequency information and the high-frequency initial estimation figure, and the compressed low-resolution image restoration method can solve the super-resolution restoration problem of the low-resolution image with the compression artifact.

Dynamically updating neural network systems may be implemented to generate, train, evaluate and update artificial neural network data structures used by content distribution networks. Such systems and methods described herein may include generating and training neural networks, using neural networks to perform predictive analysis and other decision-making processes within content distribution networks, evaluating the performance of neural networks, and generating and training pluralities of replacement candidate neural networks within cloud computing architectures and / or other computing environments.

Methods and devices are provided for implementing security groups in an enterprise network. The security groups include first network nodes that are subject to rules governing communications between the first network nodes and second network nodes. An indicator, referred to as a security group tag (SGT), identifies members of a security group. In some embodiments, the SGT is provided in a field of a data packet reserved for layer 3 information or a field reserved for higher layers. However, in other embodiments, the SGT is provided in a field reserved for layer 1 or layer 2. In some embodiments, the SGT is not provided in a field used by interswitch links or other network fabric devices for the purpose of making forwarding decisions.

A multi-protocol personality module enabling load / store from remote memory, remote DMA transactions, and remote interrupts, which permits enhanced performance, power utilization and functionality. In one form, the module is used as a node in a network fabric and adds a routing header to packets entering the fabric, maintains the routing header for efficient node-to-node transport, and strips the header when the packet leaves the fabric. In particular, a remote bus personality component is described. Several use cases of the Remote Bus Fabric Personality Module are disclosed: 1) memory sharing across a fabric connected set of servers; 2) the ability to access physically remote I / O devices across this fabric of connected servers; and 3) the sharing of such physically remote I / O devices, such as storage peripherals, by multiple fabric connected servers.

A mechanism for performing RDMA operations over a network fabric. Apparatus includes transaction logic to process work queue elements, and to accomplish the RDMA operations over a TCP / IP interface between first and second servers. The transaction logic has out-of-order segment range record stores and a protocol engine. The out-of-order segment range record stores maintains parameters associated with one or more out-of-order segments, the one or more out-of-order segments having been received and corresponding to one or more RDMA messages that are associated with the work queue elements. The protocol engine is coupled to the out-of-order segment range record stores and is configured to access the parameters to enable in-order completion tracking and reporting of the one or more RDMA messages.

A dissolvable nano web porous film has a network structure and a large specific surface area. The network structure is formed by stacking nano fibers that have a thickness of 10˜50,000 nm. The nano fibers are formed of a composition comprising nano fiber forming polymer, a gelling agent, a plasticizer, an optional active component and an optional additive. A method of preparing a dissolvable nano web film comprises producing a polymer melt or solution formed of a composition that comprises a nano fiber forming polymer, a gelling agent, a plasticizer, an optional active component and additives, spinning the melt or solution to form a nano web porous film having nano webs stacked in a network structure, drying the nano web porous film, and stabilizing the nano web porous film. The porous film has a 5˜5000 times higher specific surface area and a 2˜20 times faster dissolving time than a conventional film. The dissolvable nano web porous film has excellent dissolvability and solubility.

The invention discloses a preparation method for sodium alginate-acrylamide-based hydrogel. The preparation method comprises the following steps of: dissolving sodium alginate powder in deionized water; then sequentially adding an acrylamide monomer, a methylene diacrylamide cross-linking agent, ammonium persulfate and an N,N,N',N'-tetramethylethylenediamine catalyst; uniformly stirring the materials, pouring the mixture into a glass die, and heating the mixture to obtain a hydrogel; completely soaking the hydrogel in a 0.01-1 mol / L non-monovalent cation aqueous solution for 1-10 hours, wherein cations diffuse and enter in a hydrogel network structure, and induce sodium alginate to cross-link, so as to generate the high-strength and high-toughness sodium alginate-acrylamide-based hydrogel during the process. The hydrogel disclosed by the invention has the following performances: the highest tensile strength can achieve 1 MPa, and the highest tensile elasticity modulus can achieve 250 KPa; a loading-unloading test is performed on the hydrogel, and when the great tensile multiple before unloading is 8, the highest dissipated energy can achieve 2180 KJ / m<3>.

A SW (70) receives an Ethernet® signal from an outside of areas E and F. The SW (70) selects and outputs the obtained Ethernet® signal to any one of APs (91a to 91e) in accordance with a network structure managed by the SW (70). The AP (91a to 91e) converts the Ethernet® signal to an electrical signal type wireless LAN signal, which is in turn output to a main station (10). The main station (10) frequency-multiplexes the signal output from each of the APs (91a to 91e), and converts the signal to an optical signal, which is in turn output to sub-stations (20a and 20b) The sub-station (20a and 20b) transmits the signal transmitted from the main station (10) to a terminal in the form of a wireless radio wave. Thereby, when a plurality of communication areas are present, the accommodation capacity of an AP can be effectively utilized in each communication area.

Paths for packets traveling through a distributed network fabric are chosen using information local to the source of packets. The system allows resequencing of packets at their destination and detecting out-of-order and missing packets.

The invention discloses an image classification method based on a convolution neural network. The method comprises the following steps: constructing a deep convolution neural network; improving the deep convolution neural network; training and testing the deep convolution neural network; and optimizing the network parameter. By using the image classification method disclosed by the invention, the improvement and the optimization are respectively performed on the network structure and multiple parameters of the convolution neural network, the recognition rate of the deep convolution neural network can be effectively improved, and the accuracy of the image classification is improved.

Routes may be installed across multiple link state protocol controlled Ethernet network areas by causing ABBs to leak I-SID information advertised by BEBs a L1 network area into an L2 network area. ABBs will only leak I-SIDs for BEBs where it is the closest ABB for that BEB. Where another ABB on the L2 network also leaks the same I-SID into the L2 network area from another L1 network area, the I-SID is of multi-area interest. ABBs will advertise I-SIDs that are common to the L1 and L2 networks back into their respective L1 network. Within each L1 and L2 network area, forwarding state will be installed between network elements advertising common interest in an ISID, so that multi-area paths may be created to span the L1 / L2 / L1 network areas. ABBs may summarize BEB multicast trees such that the set of trees for a given I-SID transiting the ABB is condensed into a common tree. The L2 network may further be implemented as a second layer implemented using a L1 / L2 / L1 network structure, so that the L1 / L2 / L1 network structure may recurse an arbitrary number of times.

An enhanced method and system for the classification of a target web page and the description of a set of web pages web pages utilizing virtual documents, in which a virtual document comprises extended anchortext extracted from each of a plurality of web pages that includes at least one hyperlink citing each target web page.

There is provided an information processing apparatus which may be applied to an information processing apparatus being connected, for example, to an information network such as Internet for transmitting and receiving various pieces of information in order to make access to such network through a terminal of simplified structure such as a telephone terminal. A code for uniquely identifying, for example, a telephone terminal and a computer terminal may be formed with the codes corresponding to operators of a telephone terminal.

The invention discloses a block chain bottom consensus mechanism and a block chain system based on the consensus mechanism. The block chain bottom consensus mechanism comprises a consensus network structure, participants, a first protocol, a second protocol, and a third protocol; the consensus network structure is used for providing a carrier for transmitting and synchronizing data; the participants are used for participating in data storage and evaluation in the consensus network structure; the first protocol is used for voting for specific participants based on the rotate account sub-protocol and generating a new block when equal probability rotating; the second protocol is used for providing group voting or estimate for all participants holding tokens or digital shares based on the group estimate consensus sub-protocol and generating an extra new block; and the third protocol is used for connecting with each consensus mechanism network based on the gear consensus routing sub-protocol. According to the invention, the characteristics of the block chain data structure and the point to point network communication are combined to realize the secure, efficient, decentralized and application scenario flexible data synchronization consensus.

A storage system comprises a plurality of storage arrays, a plurality of controllers selectively coupled to ones of the storage arrays, a network fabric coupling the plurality of controllers, and a processor. The processor is capable of moving control and / or data transfer functions related to a group of elements in the plurality of storage arrays between controllers or storage arrays by transferred management information rather than data.

An electronic fund transfer system is disclosed wherein commissions for point-of-sale purchase transactions are determined and distributed to members of an organizational network for promoting use of a non-cash payment device that is tangible for effecting purchase transactions. Purchase data is electronically transmitted from the point-of-sale to an automated clearing house for effecting fund transfers to prepare the purchase, and also to a commission management system. The commission management system maintains an electronic database reflecting the structure of the organizational network, and further determines the distribution of purchase commissions according to a predetermined schedule of proportions relating to the position and business volume of each member of the network. The commission management system further comprises a payment processor for effecting commission payments to the members or to a designated sponsor on a scheduled basis.

The invention relates to an image segmentation method based on a multi-supervision full-convolution neural network. According to the invention, further improvement is carried out based on the full convolution neural network, so a novel network structure is provided. The network structure has three edge output layers with supervision which are capable of guiding a network to learn multi-scale features and allowing the network to acquire local features and global features of images. Meanwhile, in order to keep context information in the images, in the upper sampling parts of the network, upper sampling is performed on output feature images by sue of multiple feature channels. Finally, a fusion layer with the weight is used for fusing the classification results of the multiple edge output layers, so the final image segmentation result is obtained. The method is characterized by high segmentation precision and quick segmentation speed. In osteosarcoma CT data segmentation, the DSC coefficient of the acquired segmentation result by use of the provided algorithm reaches 86.88%, which is higher than that of a traditional FCN algorithm.

The invention discloses a routing method for the wireless sensor network with efficient energy, which is suitable for the layered sensor network structure. The routing method is composed of initialization, cluster building, adjacent clusters routing and routing maintenance, wherein, an initialization process of the protocol makes a Sink node obtain a topology and network average energy of the sensor network, and each node obtains hop counts from the node to the Sink node; in the stage of the cluster building, a repeated division method is used to divide sensor network clusters, the divided clusters are even, and a leader cluster node is undertaken by nodes with higher residual energy; the adjacent clusters routing uses an ant colony algorithm to determine the probability of using a link to send information according to the link pheromone concentration, and the link pheromone concentration is increased with the information transmission on the link and is reduced with the time going; and the routing maintenance stage is responsible for updating link pheromone concentration, and makes the nodes inside the cluster with higher residual energy undertake the leader cluster in turn. The routing method can reduce the consumption of the network total energy, can balance the consumption of the node energy and can prolong the network life cycle.

The invention discloses an image significance detection method which uses confrontation training to generate a convolution neural network model, which belongs to the field of computer vision and image processing. The method comprises the steps of data preprocessing, network structure, suitable parameter selecting, and training with a random gradient descending method and an impulse unit. According to data preprocessing, a large amount of collected data and labels are preprocessed. According to network structure, a network structure and a specific kernel function are designed. Suitable parameters including learning rate, a momentum factor and the number of images inserted into the network each time are selected. The random gradient descending method and the impulse unit are used for training to reduce the possibility of network over-fitting. According to the invention, a significance map can be accurately acquired.

The invention relates to organic-inorganic all-solid-state composite electrolyte as well as preparation and application methods thereof, belonging to the field of lithium ion batteries. According to the organic-inorganic all-solid-state composite electrolyte, a highly-ordered three-dimensional connection network skeleton is formed by an inorganic fast lithium ion conductor, and a three-dimensional connection network is filled with a polymer and lithium salt. The organic-inorganic all-solid-state composite electrolyte which is flexible and has a controllable three-dimensional connection network structure is prepared. The electrolyte is high in lithium ion conductivity, wide in electrochemical window, good in mechanical property and stable to lithium metal. A lithium ion secondary battery assembled by a composite electrolyte membrane prepared by the method is high in capacity, stable in cycle performance, low in interface impedance and good in interface stability.