15703 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.

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.

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.

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.

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.

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.

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 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.

The present solution is directed towards systems and methods to more efficiently control a flow of a data stream traversing at least one intermediary on a network between a client and a server. A sender transmits a first message, comprising a first value of a bandwidth between the first intermediary and a second intermediary determined by the sender, to a first intermediary. The first intermediary establishes a next value of the bandwidth between the first intermediary and the second intermediary. The sender receives from the first intermediary responsive to the first message a second message comprising the established next value of the bandwidth between the first intermediary and the second intermediary. A data transfer manager of the sender, responsive to the second message determines a size of a portion of data queued for transmission to transmit to the first intermediary and a time for transmitting the portion of data queued.

The invention provides a method for preparing a graphene-carbon nano tube hybrid composite and relates to a method for preparing a functional high molecular material and a device thereof. The method comprises the following steps of: carrying out the stirring and sonic oscillation treatment on graphene and carbon nano tubes to preform an entangled network structure, thoroughly mixing the entangled network structure and polymer particles, and thus obtaining an uniformly-mixed system in which the graphene-carbon nano tube network is coated on the surfaces of the polymer particles after removing the solvent; and putting the uniformly-mixed system in a mould, hot-compacting, and obtaining the graphene-carbon nano tube hybrid composite after cooling and demoulding. By mingling the graphene and the carbon nano tubes in advance to form the communicated network structure, the method realizes the advantage complementation of the graphene and carbon nano tube structures and enables the hybrid composite to have favorable electric conduction and heat conduction properties. The method can be widely used in the fields such as aviation, transportation and communication, electronic industry, civil facilities, construction, chemical industry and the like, can be produced in the industrial scale, and has the advantages of low cost and environmental friendliness.