51 results about "Artificial networks" patented technology

A digital circuit element of a two dimensional dynamic adaptive neural network array (DANNA) may comprise a neuron/synapse select input functional to select the digital circuit element to function as one of a neuron and a synapse. In one embodiment of a DANNA array of such digital circuit elements, a destination neuron may be connected to a first neuron by a first synapse in one dimension, a second destination neuron may be connected to the first neuron by a second synapse in a second dimension and, optionally, a third destination neuron may be connected to the first neuron by a third synapse thus forming multiple levels of neuron and synapse digital circuit elements. In one embodiment, multiples of eight inputs may be selectively received by the digital circuit element selectively functioning as one of a neuron and a synapse. The dynamic adaptive neural network array (DANNA) may implement long-term potentiation or depression to facilitate learning through the use of an affective system and random selection of input events.

The embodiment of the invention discloses a paging parameter self-optimizing method in a long term evolution network and a system thereof. The method comprises that: an eNB monitors and counts paging loads and paging performance; the counting result of the paging loads and the paging performance obtained via counting is reported to an OAM system by the eNB according to the preset period or based on the preset event trigger; the OAM system confirms correction of paging parameters and the paging parameters needed to be adjusted according to the counting result of the paging loads and the paging performance; if the paging parameters are needed to be corrected, the OAM system adopts the preset optimization algorithm to calculate adjustment values of the paging parameters needed to be adjusted and sends a paging parameter adjustment notification message to the eNB; and the eNB adjusts the paging parameters needed to be adjusted according to the paging parameter adjustment notification message and notifies the adjusted paging parameters to a user terminal UE. Paging parameter self-optimization in the LTE network can be realized by the embodiment of the invention so that the network paging performance is effectively enhanced and artificial network optimization cost is reduced.

The invention discloses an electromagnetic compatibility testing load device and a testing system of a new energy automobile electric driving system. The load device comprises a load motor, a load motor controller, a transmission device, a torque rotating speed sensor, a power supply, a first photoelectric signal converter, a second photoelectric signal converter, a monitoring device, a DC high voltage filter and a screening box; the testing system comprises the load device, an electric driving system, a low-voltage artificial network, a high-voltage artificial network, a high-voltage DC source, a DC filter, a testing table and an anechoic chamber. The load device is movable, capable of realizing high-rotating speed large torque and used for performing electromagnetic compatibility testingof the large power motor, thereby preventing an operation of importing the radiation interference from the external and preventing the load device from producing radiation interference on the external testing environment; furthermore, the testing system takes electricity from the AC power grid, and the mechanical energy of the tested motor is converted into the electric energy through the load device, and then is merged into the AC power grid, thereby realizing the cyclic utilization of the energy and reducing the experimental consumption.

The invention relates to an abnormal network connection detection method based on deep learning. The method comprises the following steps of extracting a network connection identification field for each network flow record, and aggregating all network flow records according to the network connection identification fields; constructing a network connection model based on a deep neural network; constructing an abnormal network connection detector, using the output of the network connection model as the input, and synchronously training with the network connection model in order to obtain a detection result for network connection; utilizing a data set to carry out parameter adjustment optimization and false alarm control on the network connection model and the abnormal network connection detector, and if an expected effect is achieved, finishing training and storing network parameters and structures; and inputting network connection records to be detected into a combined model of the trained network connection model and the abnormal network connection detector, and outputting the abnormal network connection records. According to the method, the abnormal network connection can be foundout, and the method does not depend on an artificial network connection model.

The invention discloses a remote-sensing estimation method for chlorophyll of apple leaves. The remote-sensing estimation method comprises synchronously acquiring hyperspectral reflectivity and corresponding chlorophyll relative content of the apple leaves by virtue of an SVC HR-1024i type hyper-spectrometer and an SPAD-502 chlorophyll meter, carrying out relevant analysis on original spectral reflectivity and a first-order derivative derivative spectrum so as to extract spectrographic red edge parameters of the apple leaves, and optimizing an artificial neural network by virtue of a traditional univariate regression algorithm, a BP neural network and a radial basis function network, and establishing a chlorophyll content inversion model. Compared with a traditional univariate model, the regression precision of an artificial network model is obviously increased, and the radial basis function network is high in study speed and precision and has a relatively reliable fitting result; and the artificial network model is the chlorophyll content inversion model which is worthy of being popularized.

The invention discloses a geometrical feature filter method of time-frequency composite overlapping signal, which is characterized by the following: utilizing artificial nerve network as non-linear filter; exercising the weight coefficient of artificial nerve network; limiting the exercising method through general BP algorism; regarding signal time-region waveshape as geometrical image or curve to realize non-linear filter for artificial network.

The invention relates to application of capacity load architecture of a cascaded failure of a complex network. On the basis of the complex network theory, loads of crashed nodes are allocated by considering heterogeneity of each node at a network structure. The node with the high degree has strong data processing capability; the node with the low clustering coefficient is able to play a big role in a low-resistance cascaded failure. More loads of the crashed nodes are allocated for nodes with the high degrees and low clustering; simulation at an artificial network and a real network and comparison with other models are carried out, so that effectiveness of the model is verified. Therefore, the network with the application can realize the global robustness quickly.

The invention discloses a parallel vehicle networking system based on an ACP method and a social physical information system. The parallel vehicle networking system based on an ACP method and a social physical information system includes a practical vehicle networking system and an artificial vehicle networking system, wherein physical space is formed in the practical vehicle networking system; a physical entity senses the physical entity information and transmits the physical entity information to the practical network space of the practical vehicle networking system; the artificial vehicle networking system is provided with a dynamic artificial vehicle networking model; the dynamic artificial vehicle networking model is established based on the physical entity information; the dynamic artificial vehicle networking model transmits the artificial vehicle networking information to the artificial network space; and a parallel execution mechanism is arranged between the practical network space and the artificial network space to perform information parallel interaction to perform real time optimization and updating on the dynamic artificial vehicle networking model and feedback the result which is obtained through optimization processing of computational experiments to the practical vehicle networking system, that is, a parallel execution mode is utilized between the practical vehicle networking system and the artificial vehicle networking system to realize an efficient on-line bidirectional feedback mechanism of the parallel vehicle networking system.

The present invention provides a test method of switch power high-low voltage coupling attenuation characteristics. The test is performed in a shielding chamber, and a signal source generates test signals; the test signals are output to an electromagnetic clamp after attenuation matching by a power amplifier and an attenuator, the signals are injected into the high-voltage alternating current sideof the switch power source through the electromagnetic clamp, the test signals generate an amplitude of Vinput at the high-voltage alternating current side of the switch power source; the test signals with the amplitude of Vinput are coupled and penetrate the switch power source to generate attenuation, an artificial network outputs the attenuation signals, and a receiver measures the amplitude of the signals to be Voutput; and workers subtract the Voutput from the amplitude Vinput to obtain a test result of the high-low voltage coupling attenuation characteristics of the switch power source.The test method of switch power high-low voltage coupling attenuation characteristics is simple and practicable, clear in principle, is provided with a common test device, is low in cost and high inuniversality, can effectively ensure the repeatability and the consistency of the test.

The invention relates to an electromagnetic transient modeling and calculating method for a power grid containing a multi-voltage-source converter, and belongs to the technical field of dynamic simulation of power systems, microgrids and multiple power electronics. According to the method, converter switching device trigger control and switching action processes are considered, the original topology of a converter main circuit is reserved, and converter bottom layer control, switching transient and connection coupling lines and networks between converters can be accurately simulated. Accordingto the method, segmentation processing is not carried out. The coupling of the converters on the direct-current side and the alternating-current side is completely reserved. The consistent convergence of multi-converter networking whole-system solution is guaranteed. Errors caused by approximate decoupling or artificial network segmentation are eliminated. The method is suitable for electromagnetic transient process simulation of grid connection and networking of multiple power electronic converters, such as efficient simulation and real-time simulation of detailed electromagnetic transient processes of new energy grid connection, a direct-current power grid, a flexible direct-current back-to-back near-end power grid, a microgrid containing a distributed power supply and a shipboard airborne microgrid.

The invention discloses a simulation method and device for predicting electromagnetic radiation, and the method comprises the steps: building an initial simulation model for predicting electromagnetic radiation, wherein the simulation model comprises part models of a PCB, a cable, a load end, a power supply, an artificial network, a metal ground, a test desktop and an antenna model, and determining the shape and spatial distance of each part model; defining material attributes of each part model according to a finite element calculation electromagnetism algorithm to obtain a final simulation model; based on the final simulation model, defining a frequency domain power source at the connection point of the power supply and the cable, defining an antenna model port to receive a load, and integrating an electric field on a port connection line at the load receiving position; performing mesh generation on the final simulation model according to finite element computational electromagnetism; and simulating to obtain a port voltage result of the antenna model. According to the invention, the influence of the test antenna on the test result is considered, and the simulation application with the simulation being more suitable for the actual test working condition is constructed, so that the simulation result is more accurate.

The invention relates to a label propagation natural heuristic-based dynamic network community structure identification method, and belongs to the field of artificial intelligence and complex networks. According to the invention, the label propagation algorithm is used for initializing network communities and restricting the conditions of the variation process, so that the detection efficiency andeffectiveness can be further improved; Through genetic manipulation, diversity is increased on the premise that a community structure is maintained; And a particle swarm algorithm is used to avoid falling into local optimum, and a global optimum solution is obtained. And finally, according to the experimental results of testing on the artificial network and the real network, the node tags can beupdated according to the node degrees through the improvement of the tag propagation algorithm. And the nodes with high degrees have greater influence on surrounding nodes, so that the problem that the iteration result is unstable is solved. And one-time synchronous updating operation is carried out by using the abrupt change operation and the particles with the random numbers smaller than the abrupt change rate. Not only is a good community structure maintained, but also diversity of particle positions is increased, and local optimum is avoided.

The invention belongs to the field of communication, control and computing technologies and relates to a novel network system architecture. The ACP method-based parallel network architecture is characterized by comprising the following steps of A, modeling: establishing an artificial network model, perfecting the artificial network model, and increasing interference parameters for the model at special positions; B, conducting the calculation experiment: conducting the calculation experiment on historical data and storing to facilitate the comparative invocation of the actual network; C, parallel execution: adjusting the state of the artificial network and optimizing the network while correcting the system state of the actual network at the same time.

The invention provides a community discovery method for phylogenetic transplantation partitions, which comprises the following steps: S1, acquiring data information of a current community, and taking the data information as community information to be processed; the preprocessing comprises the following steps: processing all states as variables, and constructing a time sequence dynamic observation model of systematic evolution; s2, preprocessing the to-be-processed community information obtained in the step S1; s3, constructing an error function, and then performing minimization processing on a quadratic form of the error function; s4, solving the gradient of the error function, and carrying out iteration according to the gradient direction; and S5, obtaining data information of the subarea communities. Compared with a FaceNet method, an SBM + MLE method, a CLBM method and a PisCES method, the PPPM model provided by the invention has the advantages that the accuracy is improved by 5% and 3% on an artificial network and a real network respectively, so that the provided PPPM model has robustness, is reasonable and effective, and can also be applied to the field of common social network community discovery.

The invention discloses an online voltage drop source identification system and method. The online voltage drop source identification system comprises an electric system simulating calculation program module, a training database and an artificial neural network program module, wherein the electric system simulating calculation program module is used for carrying out simulating calculation on different bus faults and obtaining the voltage simulating calculation result of each node, the training database is connected with the electric system simulating calculation program module and used for storing voltage effective values of all buses in an electric system under different fault conditions, the voltage effective values are worked out through an electric system simulating calculation program, and the artificial neural network program module is connected with the training database and used for carrying out artificial network training through data in the training database. According to the online voltage drop source identification system and method, the position where a voltage drop source is triggered can be rapidly worked out according to the bus voltage monitoring result of electric energy quality monitoring devices arranged on part of the buses, and a basis is provided for dispatching operators to handle the faults.

The invention discloses an automatic generation method of an inter-domain route protocol traffic engineering link, which includes: step S102, look up link status database of interior gateway protocol and acquire all domain border routers IDs; step S104, based on the domain border routers ID, construct and generate a search request of the inter-domain route protocol traffic engineering link among inter-domain routers, in order to search and accept search results; step S106, generate the inter-domain route protocol traffic engineering link, according to the search results. In addition, the invention also discloses an automatic maintenance method of the inter-domain route protocol traffic engineering link. Through the technical proposal thereof, the invention can feedback the network topotactic transformation results dynamically and quickly, effectively improve the network resources utilization rate, avoid the complex artificial operation, reduces the occurrence probability of artificial network failures and increases the network stability, by the realization of the automatic link generation.

Various embodiments of the present technology generally relate to robotic devices and artificial intelligence. More specifically, some embodiments relate to an artificial neural network training method that does not require extensive training data or time expenditure. The few-shot training model disclosed herein includes attempting to pick up items and, in response to a failed pick up attempt, transferring and generalizing information to similar regions to improve probability of success in future attempts. In some implementations, the training method is used to robotic device for picking items from a bin and perturbing items in a bin. When no picking strategies with high probability of success exist, the robotic device may perturb the contents of the bin to create new available pick-up points. In some implementations, the device may include one or more Computer-vision systems.

The invention provides a dynamic community discovery method for a phylogenetic transplantation partition sequential network. The method comprises the following steps: S0, defining a social network to obtain a sequential social network of the social network; the method comprises the following steps: S1, collecting data information of a current community, and taking the data information as to-be-processed community information; s2, preprocessing the to-be-processed community information obtained in the step S1; s3, constructing an error function, and then performing minimization processing on a quadratic form of the error function; judging the reliability range; s4, solving the gradient of the error function, and carrying out iteration according to the gradient direction; and S5, obtaining data information of the subregion communities. Compared with a FaceNet method, an SBM + MLE method, a CLBM method and a PisCES method, the PPPM model provided by the invention has the advantages that the accuracy is improved by 5% and 3% on an artificial network and a real network respectively, so that the provided PPPM model has robustness, is reasonable and effective, and can also be applied to the field of common social network community discovery.

The invention discloses an alloy grade identification method based on an artificial neural network. The method comprises the steps of (1) establishment of the artificial neural network in which the proper artificial neural network which includes the type and layer number of the neural network, the number of neurons in each layer, a transfer function and a training function is established according to the size of a grade library; (2) training of the artificial neural network in which sample input includes contents of key elements, sample output includes identification codes of grades, and the trained artificial network is stored and kept standby; and (3) simulation of the artificial neural network in which actual measurement data of detected samples is input to the trained neural network, integers are taken from the output data of the neural network proximately, the taken values represent the identification codes of grades of the detected samples, and the grades of the detected samples are reported according to the identification codes of the grades.