68 results about "Hopfield network" patented technology

To provide a semiconductor device with a small circuit size and low power consumption or an electronic device including the semiconductor device and compressing a large volume of image data. A semiconductor device of a Hopfield neural network is formed using neuron circuits and synapse circuits. The synapse circuit includes an analog memory and a writing control circuit, and the writing control circuit is formed using a transistor including an oxide semiconductor in a channel formation region. Thus, data retention lifetime of the analog memory can be extended and refresh operation for data retention can be omitted, so that power consumption of the semiconductor device can be reduced. The semiconductor device enables judgement whether learned image data and arbitrary image data match, are similar, or mismatch by comparing video data. Thus, motion compensation prediction, which is one of data compression methods, can be employed for image data.

The invention provides an analog circuit fault diagnosis method based on wavelet packet analysis and the Hopfield network. The method includes data obtaining, feature extraction and fault classification, wherein data obtaining includes performing data sampling for output response of an analog circuit respectively through simulation program with integrated circuit emphasis (SPICE) simulation and a data collection plate connected at a practical circuit terminal so as to obtain an ideal output response data set and an actually-measured output response data set; feature extraction includes performing wavelet packet decomposition with ideal circuit output response and actually-measured output response respectively serving as a training data set and a test data set, and leading energy values obtained by decomposed wavelet coefficient through energy calculating to form feature vectors of corresponding faults; and fault classification includes leading the feature vectors of all samples to be subjected to Hopfield coding and then submitting the coded feature vectors to the Hopfield network to achieve accurate and fast fault classification. The analog circuit fault diagnosis method is good in fault feature pretreatment effect aiming at hard faults with weak amplitude response and soft faults with large amplitude response, and the newly defined energy function and the newly defined coding rule are remarkable in influence on fault diagnosis accuracy of the analog circuit.

The present invention relates to the digital image processing technology, especially to a water level identification method based on binary coding character staff gauge and image processing. The water level identification method based on the binary coding character staff gauge and the image processing comprises: extracting the binary coding character staff gauge image key pixel, performing binary coding character staff gauge tilt correction through adoption of the Radon conversion algorithm; determining the left, the right, the upper and the lower edges of the binary coding character staff gauge; extracting the binary coding character staff gauge scale line; performing the location and the segment of the binary coding character representing the staff gauge measuring range; performing the binary coding character staff gauge measuring range identification through combination of the Hopfield nerve network and the template matching; and realizing the resolving of the water level value through adoption of the mathematical relationship of the binary coding character staff gauge scale line and the binary coding character staff gauge measuring range. The water level identification method based on the binary coding character staff gauge and the image processing is safe, efficient and small in investment.

The embodiment of the invention relates to a distribution method of network flow. The distribution method comprises the following steps of: establishing a plurality of optional routing paths between a source node and a destination node of a network, and initially distributing the network flow between the source node and the destination node to one or more optional routing paths; according to the network flow initially distributed to the optional routing paths, determining the initial optimal routing path in the optional routing paths by a hopfield neural network algorithm, and distributing the network flow between the source node and the destination node to the initial optimal routing path; and according to the network flow initially distributed to the optional routing paths and the network flow distributed to the initial optimal routing path, adjusting the network flow distributed to the optional routing paths by an FD flow deviation algorithm until the network transmission time delay meets the predetermined requirements. In the invention, the multiple optional routings are established between the source node and the destination node, and the hopfield neural network algorithm and the FD algorithm are combined for adjusting the service load of each link in the network, thereby adjusting the flow distribution and optimizing the network transmission time delay.

In view of the shortcomings of the traditional three-dimensional Hopfield neural network, the invention presents a four-dimensional Hopfield neural network image encryption method based on quantum Fourier transformation. First, a hyper-chaotic sequence of a four-dimensional Hopfield neural network is generated with the aid of plaintext information and an input key. Then, a quantum chaotic sequenceis generated with the aid of a mapping NCML network through quantum Fourier transformation to carry out secondary encryption in order to improve the shortcoming of the traditional quantum Fourier without chaotic characteristic. Finally, Arnlod scrambling is introduced to get a final ciphertext. The simulation experiment shows that the algorithm not only can effectively resist statistical featureattack and differential attack, but also can greatly improve the shortcomings of the traditional three-dimensional Hopfield neural network such as low dynamic complexity and small Lyapunov index, andachieve a good encryption effect.

This invention relates to one SoC software and hardware division method based on discrete Hopfield neutral network, which comprises the following steps: adopting pattern description method to divide the software and hardware problem into one detail combination optimization problem to introduce SoC division problem new module; then according to the division property, re-defining discrete Hopfield neural network element, energy function, operation equation and parameters; dividing the discrete Hopfield network as division formula on SoC chip functions.

The invention discloses an MMSE-BDFE (Minimum Mean Square Error-Blind Decision Feedback Equalizer) multi-user detection system based on a neural network. The MMSE-BDFE multi-user detection system is characterized by comprising a receiver, a noise adder, a sampler, a filter, a channel estimating unit and a neural network signal detection processing unit. A working method of the MMSE-BDFE multi-user detection system comprises the steps of: receiving a signal, obtaining noise, sampling and filtering, processing a data function and outputting a signal. The MMSE-BDFE multi-user detection system has the advantages that the structure is simple, the convenience is brought for the operation, the computing complexity of an MMSE-BDFE is reduced, the MMSE-BDFE is optimized to be crossed with a neural network, multi-site interference is inhibited, communication quality and system stability are improved; and an optimization program of multi-user detection corresponds to an energy function of a Hopfield neural network, thus the instantaneity is improved.

The embodiment of the invention discloses an improved Hopfield neural network based recognition method for a clothes logo, applies to the field of computers, and aims at solving the problem of extremely low manual recognition efficiency of goods logos in the prior art. The method comprises the steps of extracting characteristics of an original image through wavelet transformation; building a Hopfield neural network model by using the characteristics as the target mode. The method is applied to logo recognition.

The invention relates to a clustering multi-hop routing method based on a maximum and minimum distance method. A maximum and minimum distance method is adopted to select a cluster center, re-clustering is carried out according to the cluster center, the selection of a cluster head is carried out according to a maximum weight principle, a new cluster is formed according to distance between a node and the cluster head, the insufficiency of random selection of a LEACH (Low Energy Adaptive Clustering Hierarchy) protocol cluster head is overcome, so that the energy consumption of a network is evenly consumed on each node. A link with a shortest communication path is generated between the cluster head and Sink by adopting a continuous Hopfield neural network, the link is optimized to form a multi-hop tree cluster type link which takes the Sink as a center, and energy consumption generated by long-distance communication caused by adopting a single-hop way is lowered. The deficiencies of the cluster head selection and a node single-hop mechanism in the LEACH protocol under the energy heterogeneous environment of a wireless sensor network can be eliminated, and aspects on network stable phase prolonging and energy balance are obviously improved than a LEACH algorithm.

The invention relates to a reciprocating type compressor indicator diagram testing device and a fault diagnosis method, and belongs to the technical field of error diagnosis of a reciprocating type compressor. According to the reciprocating type compressor indicator diagram testing device and the fault diagnosis method disclosed by the invention, an indicator diagram of the reciprocating type compressor is identified by adopting discrete type Hopfield nerve network, and meanwhile, an associative memory function of the discrete type Hopfield nerve network is utilized, so that the indicator diagram of the reciprocating type compressor is accurately measured and identified by utilizing the function, and a satisfactory effect is achieved; and moreover, the calculation convergence speed is high.

The invention discloses a direct current master device fault diagnosis method based on a hybrid neural network. The method includes the following steps that firstly, associated data needed for device fault diagnosis are acquired, wherein the associated data comprise source data and real-time data, and the source data include offline experimental data, dot experimental data, online monitoring data and historical data composed of various polling data; secondly, information fusion is conducted on the associated data through a neural network; thirdly, a particle swarm optimization algorithm, a Hopfield network and a BP network are combined, the hybrid neutral network is designed, the associated data obtained after information fusion in the second step are predicted, and then the prediction state of a direct current master device is acquired; fourthly, the prediction state corresponds to the original state of the direct current master device and is shown in different modes or/and forms, wherein the original state is the historical state shown by the source data. By means of the method, the overhaul efficiency of the fault device and the running reliability of a power grid are improved.

The invention discloses a masonry beam deformation monitoring system and monitoring and forewarning method based on a Hopfield neural network. The masonry beam deformation monitoring system comprisesa grating optical fiber sensor arranged at each monitoring point in a roadway and an optical fiber grating demodulator arranged on a well and in communication connection with the grating optical fibersensors through a downhole monitoring system platform. The signal output end of the optical fiber grating demodulator is connected with a Hopfield neural network processing system, and the Hopfield neural network processing system and a storage searching server conduct information interaction. By adopting the masonry beam deformation monitoring system and monitoring and forewarning method, multi-point and multi-parameter monitoring is achieved; the grating optical fiber sensor are assembled to collect data, mining optical fibers are adopted for transmission, so that the capability of resisting electromagnetic interference is high, and operation is stable; and the data are processed based on the Hopfield neural network processing system, and a data platform is shared, so that parameters can be optimized, the forewarning and alarming threshold value can be further adjusted according to the deformation situation of the downhole roadway, and thus the efficacy of the monitoring system is improved greatly.

The invention provides a method and a system for hazily and comprehensively evaluating the power efficiency of an enterprise user by using a Hopfield neural network method. The system comprises a basic data acquisition platform consisting of a power detector and a communication server, and an efficiency evaluation platform. The method comprises the following steps that the power detector communicates all acquired data and data analyzed by the detector with the communication server through an RS485 interface; the data is transmitted to a power efficiency evaluation platform of a company through an inner-enterprise local area network, and software in the efficiency evaluation platform classifies, analyzes and counts the received data according to an efficiency evaluation system to obtain each factor value of an efficiency index system; and factors in each factor group are evaluated by the Hopfield neural network method. By the method and the system, various efficiency influence factors of the enterprise can be determined, the requirement on objectiveness of the evaluation process is met by fully utilizing the learning capacity and adaptive ability of artificial neuron, and the evaluation system is more objective and practical.

The invention discloses a Hopfield neural network-based server energy-saving device for a cloud data center. The device comprises a data storage part, a control part and an energy-saving strategy training part, wherein the data storage part is used for storing monitoring data and energy-saving strategy information of a server group; the control part is responsible for service control, wherein the service control comprises generation and acquisition of the monitoring data, and matching and implementation of an energy-saving strategy; the energy-saving strategy training part is used for training a Hopfield neural network based on the monitoring data to generate the energy-saving strategy information. The invention also discloses a corresponding method. According to the device and the method, the problems that most energy-saving strategies are single in setting and are not accurate and reasonable enough, and the energy consumption of the data center cannot be adjusted very well are effectively solved.

The invention discloses a customer classification method and system based on data mining. The method comprises the following steps of 1, determining target customer objects and target data; 2, collecting the data, and preprocessing the customer data by means of data cleaning, data integration, data conversion, attribute reduction and principal component analysis; 3, adopting and combining a K-means algorithm and a Hopfield neural network algorithm to mine the customer data, and obtaining classification results of different customer groups; 4, analyzing and evaluating the classification resultsof the step 3, and if the classification results are optimal, displaying a mining result, otherwise executing the step 3 again. By means of the customer classification method and system based on datamining, the customer data can be effectively and accurately classified so that enterprises can timely and accurately master customer resources and changing tendencies, and the customer resources areeffectively managed.

The invention relates to a robust multi-user detection method based on a quantum Hopfield neural network and a quantum fish swarm algorithm in an impact noise environment. The method comprises the steps that a robust multi-user detection model is established; the quantum Hopfield neural network is activated, and a second-best solution is generated; a quantum fish swarm is initialized; an evolution rule of the quantum artificial fish swarm algorithm is used for carrying out evolution on a population; according to a food concentration function, food concentration values are computed for all new positions; and an obtained global optimum position is sending data for detecting a plurality of users, and detecting results are output. Robust multi-user detection in the strong impact noise environment is well achieved, the designed quantum Hopfield neural network and the quantum fish swarm algorithm are used as evolution strategies, and the designed method has the advantages of being high in convergence rate and high in convergence precision.

The invention discloses a phase shift keying signal blind detection method based on a plural discrete full-feedback neural network. In the method, according to a principle of decreasing energy function of the plural discrete full-feedback neural network, a Hermitian weight matrix capable of directly detecting phase shift keying signals is constructed, so that each multi-phase shift keying (MPSK) signal centralized constellation signal point is a stable equilibrium point of a Hopfield neural network; therefore, the blind detection of the MPSK signals is realized. The method can realize computing targets by only needing extremely short received data, and can be applied to statistic meaningless occasions. The search space is narrowed, the difficulty is reduced, the search time is obviously superior to that of other blind detection algorithms, and the system performance is correspondingly improved.

A compact dual function Random Number Generator and Stream Cipher Generator includes a Crypto-engine has a controller for controlling the engine to operate in one or other of its functions. The Crypto-engine incorporates a plurality of clipped Hopfield Neural Network pairs.

The invention discloses a blind detection algorithm based on M2M communication frequency spectrum sharing and coexistence. The blind detection algorithm comprises the following steps: step SS1: constructing a traditional user overcomplete model and an M2M communication sparse model; step SS2: performing convex optimization solution on an M2M transmission signal; step SS3: constructing a traditional user receiving data matrix; step SS4: performing singular value decomposition on the receiving data matrix in the step SS3; step SS5: setting a weight matrix; and step SS6: selecting an activation function of the Hopfield neural network, and performing a Hopfield neural network iterative operation. According to the blind detection algorithm disclosed by the invention, the restoration of a traditional user in the M2M communication is achieved by using the blind detection of the Hopfield neural network, a state equation is iterated according to the traditional user overcomplete model and the M2M communication sparse model by using an M2M device transmission signal obtained by convex optimization, and during each iteration, the Hopfield neural network is entered to verify that the error rate of the blind detection algorithm disclosed by the invention is better than that of the method for restoring the traditional user method assuming that a channel is known under the same conditions bymeans of simulation.

The invention discloses an urban water supply pipe network optimizing method. The urban water supply pipe network optimizing method comprises the steps of establishing a water supply main pipe optimizing model by the adoption of the annular worth method, then establishing a motion equation of an energy function of a continuous Hopfield neural network model and optimization variables of the continuous Hopfield neural network model for water supply main pipe optimization by the utilization of the Lagrange function, and at last, solving the model established at the first step by the utilization of the mixed calculation method of the Hopfield neural network algorithm and the simulated annealing algorithm. The Hopfield neural network algorithm is parallel computation, exponential explosion of the calculated amount of the Hopfield neural network algorithm cannot occur though the number of dimensions increases, and the urban water supply pipe network optimizing method is particular effective to high-speed calculation of water pipe network optimizing problems.

This invention relates to one RTOS consumption optimization method based on discrete Hopfield neural network, which comprises the following steps: by use of pattern description method to divide the RTOS consumption relative soft and hardware into one detail combination optimization problem to introduce one new module of RTOS consumption division problem; then according to the division property re-defining discrete Hopfield neural network element, energy function, operation equation parameters to use the neural network as RTOS consumption division optimization to realize the function division to reduce RTOS power consumption.