50 results about "Hierarchical neural network" patented technology

The invention discloses a power distribution automation terminal fault detection method and system, and relates to the technical field of fault detection, and the key points of the technical scheme are that multi-modal data of a power distribution automation terminal are collected in real time through a multi-modal sensor array, and the multi-modal data comprise operation environment parameters, electrical quantity parameters and network topology data; constructing a dynamic coupling factor matrix by quantifying a nonlinear association relationship between the operating environment parameters and the electrical quantity parameters; adjusting a fault detection threshold in combination with the dynamic coupling factor matrix and the network topology data; and based on the adjusted fault detection threshold, performing fault diagnosis on the multi-modal data by using a hierarchical neural network model to generate a preliminary fault positioning result. According to the method, synchronous optimization of fault detection sensitivity and specificity in a complex distribution network environment is realized, and the problems of environment-electrical coupling failure, topology tracking lag, fault positioning fuzziness and the like in the prior art are systematically solved.

The invention discloses an audio data compression method and device, electronic equipment and a storage medium, relates to the technical field of voice processing, can be applied to financial science and technology and medical health business scenarios, and comprises the following steps: obtaining target audio data to be compressed; and inputting the target audio data into an audio coding and decoding quantization compression model to obtain a reconstructed audio signal corresponding to the target audio data, the audio coding and decoding quantization compression model being obtained through audio adversarial training. In the model, feature extraction and compression can be performed on target audio data by using a hierarchical neural network of an encoder to obtain a low-dimensional potential feature vector; a residual vector quantization module performs discrete quantization processing on the low-dimensional potential feature vector through a multi-layer cascaded codebook to obtain a discrete quantization code; and performing audio waveform reduction processing on the discrete quantization code by using a decoder to obtain a reconstructed audio signal corresponding to the target audio data. According to the invention, audio high-fidelity compression can be realized, and audio reconstruction tone quality is improved.

A closed-loop neuromodulation system, including an electrode array that is implantable to a brain of a subject, analog front-end device (AFD) for selectively selecting and reading a plurality of channels from electrode array, a finite impulse response (FIR) filter for selectively filtering signals from the AFD, a feature extraction engine (FEE) operatively connected to the FIR filter, configured to selectively extract features from signals provided by the FIR filter, a tree-structured hierarchical neural network classifier for detecting disease symptoms, and a multi-channel stimulator having high-voltage (HV) drivers operatively connectable to the electrode array.

The invention discloses a wireless communication network intelligent optimization method and system based on neuron collaboration, and relates to the technical field of wireless communication. The method comprises the following steps: mapping a communication node into a bottom layer sensing neuron and constructing a neural state variable set; constructing a node neural situation function based on the variables; when the local threshold value is exceeded, excitation pulses are generated and uploaded to middle-layer convergence neurons; the middle layer carries out pulse space aggregation and extreme value search, and outputs a selection strategy and an adjustment strategy; and reporting to a top layer to execute whole network neural synchronization index analysis, and optimizing the wireless communication network. The technical problems of low spectrum resource utilization rate and unstable network performance caused by the fact that a traditional wireless communication network cannot realize high-efficiency spectrum allocation and dynamic topology reconstruction under user mobility change are solved, and the purposes of realizing local quick response and global collaborative optimization by constructing a layered neural network and improving the network performance are achieved. And the spectrum resource utilization rate is improved, and the network dynamic adaptive capability is enhanced, so that the user service quality is guaranteed.

The invention discloses a time-varying and time-lag multi-CSTR system adaptive control method and system based on a neural network, and the method comprises the steps: collecting key operation parameters of a multi-CSTR system in real time, constructing a time-varying and time-lag dynamic model through a recurrent neural network, and updating a time-lag parameter estimation value. And designing a hierarchical neural network structure comprising an upper global coupling model and a lower local compensation model, designing an adaptive algorithm based on a stability theory to adjust the weight of the neural network, and generating a control instruction to drive an execution mechanism. The system correspondingly comprises a data acquisition layer, a time-varying time-delay estimation module and the like. According to the scheme, the problems of time varying, time lag and coupling of the multi-CSTR system are solved, and stable operation of the system is guaranteed.

The invention relates to the technical field of data analysis, and discloses an AI-based cloud data intelligent analysis and service management system and method. The system comprises a cloud data priority division module, a resource demand prediction module, a cloud data scheduling scheme generation module and a service management dynamic adjustment module. The method comprises the steps of firstly collecting multi-source cloud data, performing data preprocessing, and then performing data priority division; secondly, introducing a hierarchical asynchronous processing architecture to construct a hierarchical neural network prediction model, and outputting a predicted value according to a data priority; establishing a resource scheduling objective function according to the predicted value, optimizing cloud data scheduling parameters in a resource scheduling process by using a multi-strategy fusion grey wolf algorithm, and generating a cloud data scheduling scheme; and finally, calculating a service management health degree, and establishing a health degree response mechanism closed-loop dynamic adjustment cloud data scheduling scheme. By analyzing and processing the cloud data, the purposes of intelligent analysis and service management are achieved, and the method is accurate and objective.

A system and method for robotic teleoperation enable a teleoperated robotic element to perform a sequence of actions based on intention estimation, eliminating the need for continuous human control. The system includes a robotic teleoperation input that receives motion inputs and gaze data from a human operator performing a robotic teleoperation task. A robotic teleoperation feature extractor analyzes and processes the motion inputs and the gaze data into sequential input data. A multi-window model assigns hierarchical prediction windows to the input data, generating windowed sequential data. A hierarchical neural network processes the windowed sequential data to determine low-level action intentions and high-level task intentions, and generates an intention estimation based on the low-level action intentions and the high-level task intentions. A hierarchical dependency model incorporates hierarchical dependent loss to refine the intention estimation.

There are provided a medical image processing device, a hierarchical neural network, a medical image processing method, and a program capable of achieving highly accurate and real-time processable region-of-interest detection and class classification. A medical image processing device acquires a medical image, extracts a first feature amount and a second feature amount having a resolution relatively higher than a resolution of the first feature amount from the medical image by processing the medical image in a feature extraction network of a hierarchical neural network, detects a region of interest included in the medical image by processing the first feature amount in a first subnetwork of the hierarchical neural network, and classifies the region of interest by processing the second feature amount in a second subnetwork of the hierarchical neural network.

The invention discloses a decision-making method for existing parameter hybrid driving, particularly relates to the field of simulated confrontation, and is used for solving the problem of group collaboration failure caused by dynamic conflicts of vision and text parameters, and solving the problem of group collaboration failure through adaptive downsampling and phase compensation when fusion of high-frequency vision parameters and low-frequency text parameters is processed. The feature counteracting effect in nonlinear fusion is reduced; meanwhile, on the basis of conjoint analysis of time-frequency energy distribution and causal association strength, a weight suppression matrix is constructed to weaken strategy contradictions, a strategy causal path is strengthened, and coordination consistency of decision instructions is guaranteed; besides, through cooperation of the hierarchical neural network and a time-space gating mechanism, dynamic fusion of an action sequence and strategy constraints is realized, and response speed and strategy consistency are improved. And finally, nonlinear oscillation of the instruction stream is inhibited in real time by means of damping control, group strategy deviation is prevented, and thus the overall decision quality is optimized.

The invention discloses a heterogeneous wind power plant power control method and system based on a hierarchical neural network. The method comprises the following steps: S1, acquiring heterogeneous wind turbine generator parameters; s2, establishing a heterogeneous wind power plant state space model according to the heterogeneous wind turbine generator parameters; s3, establishing a heterogeneous wind power plant prediction model based on a time domain convolutional network-self-attention mechanism hierarchical neural network, training the heterogeneous wind power plant prediction model through a training set constructed by heterogeneous wind turbine generator parameters, and outputting a model compensation amount to correct a heterogeneous wind power plant state space model; and S4, by taking minimization of voltage and power fluctuation of the heterogeneous wind power plant as a dual control target, on the basis of the corrected heterogeneous wind power plant state space model, through a time sequence optimization heterogeneous wind power plant power instruction, realizing dual suppression of voltage and power fluctuation. The method has the advantages of improving power prediction precision, suppressing voltage fluctuation and power oscillation and the like.

To provide a medical image processing apparatus, a hierarchical neural network, a medical image processing method, and a program configured to implement highly accurate and real-time detection and classification of a region of interest.SOLUTION: A medical image processing apparatus is configured to: acquire a medical image; process the medical image using a feature extraction network of a hierarchical neural network to extract, from the medical image, first features and second features which are relatively higher in resolution than the first features; process the first features with a first sub network of the hierarchical neural network to detect a region of interest included in the medical image; and process the second features with a second sub network of the hierarchical neural network to classify the region of interest.SELECTED DRAWING: Figure 3

The invention provides a diversified multi-hop question generation method and system based on knowledge combination sampling, and relates to the technical field of multi-hop questions.The generation method comprises the steps that context knowledge and target answers are obtained, and the target answers are obtained based on a hierarchical neural network architecture; obtaining conditional probability distribution data and an initial knowledge combination according to the context knowledge and the target answer; obtaining a plurality of diversified knowledge combinations according to the initial knowledge combination and the conditional probability distribution data through a preset sampling strategy; and generating a corresponding multi-hop question according to all the diversified knowledge combinations and the target answer through a preset multi-hop question generation model. The whole process does not need complex intermediate links and can be realized by relying on a mature neural network architecture, the manpower and hardware cost of technology landing is reduced, for example, when a multi-hop question bank is constructed in the education field, a teacher only needs to input a textbook text and knowledge points, multi-hop questions of different examination angles can be automatically generated, and the efficiency is improved. And the problem of data sparsity of the multi-hop question-answer data set is effectively relieved.

The application discloses a power tower point cloud segmentation method and system based on a layered neural network, and the method comprises the following steps: acquiring point cloud data of a power tower to be processed and a surrounding environment; inputting the point cloud data into a layered neural network; sampling the point cloud data by using a multi-stage set sampling layer; performing up-sampling on a local feature vector of each centroid obtained by sampling by using a multi-stage feature propagation layer, thereby completing feature learning of the point cloud data, obtaining a global feature vector of the entire point cloud, mapping the global feature vector of the entire point cloud into a category score vector by using a full connection layer, and converting the category score vector into a probability distribution by using a Softmax activation layer, so as to obtain the probability of each point in the point cloud data belonging to different categories. The application aims to better adapt to the point cloud data features of the power tower, and realizes fast, accurate and automatic segmentation of the power tower point cloud data by using the layered neural network.

The invention discloses a fast table tennis ball rotation estimation method based on a hierarchical neural network, which comprises the following steps: constructing a table tennis ball flight path data set which comprises seven rotation types of no rotation, upward rotation, downward rotation, left upward rotation, right upward rotation, left downward rotation and right downward rotation, and each rotation type comprises multiple gears of rotation speeds; constructing a hierarchical rotation estimation network, wherein the network comprises a plurality of hybrid convolution attention modules connected in series; performing network training based on the constructed hierarchical rotation estimation network to obtain a trained hierarchical rotation estimation network; and inputting a plurality of front sampling points of the track to be measured into the trained layered rotation estimation network, and synchronously outputting the rotation type, the coarse granularity rotation speed and the fine granularity rotation speed to obtain a rotation estimation result. According to the method, convolution and a multi-head attention mechanism are fused through the hierarchical neural network, and the self-adaptive gating module and the hierarchical constraint loss function are combined, so that synchronous estimation of the rotation type and the rotation speed is realized, the recognition precision is remarkably improved, and the problem of misjudgment is solved.

The invention relates to the technical field of middleware service load prediction and resource optimization, in particular to a middleware service multi-granularity load prediction and intelligent resource pooling performance optimization method and system. The system comprises a local load feature extraction module, a global load distribution modeling unit, an adaptive resource scheduling engine and a resource utilization rate monitoring feedback loop. The load data processing efficiency is improved through a sliding window mechanism and a distributed architecture, the load prediction accuracy is enhanced by utilizing a hierarchical neural network and dynamic weight adjustment, and efficient resource allocation is realized in combination with task priority analysis and heterogeneous resource adaptation. The resource utilization rate and the task processing efficiency in the complex distributed system can be remarkably improved, and the method has wide application prospects and practical value.

The invention discloses a power-economic mapping matching method and system based on an improved hierarchical neural network model, and the method comprises the steps: inputting power data and economic data into a feature coding layer for feature extraction, and obtaining context semantic features of the power-economic data; inputting the context semantic features into an event relation modeling layer, constructing a multi-type graph structure by taking time nodes as event nodes in the graph structure, and modeling the multi-type graph structure to obtain a power-economic state sequence; performing feature matching on the electric power-economic state sequence through a preset electric power-economic feature matching mechanism of a feature matching layer to obtain a semantic similarity between the current electric power economic state and a target economic structure; and the output prediction layer constructs a task output structure according to different task requirements on the basis of semantic similarity, and obtains a final conclusion of power-economy mapping in combination with structural features and a matching result. According to the invention, the expression ability and decision-making assistance ability of the model to a complex coupling system are significantly improved.

The invention discloses a multi-scale point cloud attention defect detection method for key parts of aerospace equipment, and belongs to the technical field of intelligent detection, and the method comprises the steps: a point cloud optimization module carries out the preprocessing of an input sparse point cloud through a geometry-density bimodal optimization mechanism, and highlights the spatial distribution characteristics of a potential defect region; the point cloud up-sampling module constructs an encoder-generator-discriminator cascade architecture by means of a multi-network fusion mechanism, and realizes global point cloud construction and adversarial network generation of dense point clouds through a self-attention mechanism; the defect detection module realizes recognition and positioning of typical defect areas such as cracks, pits and deformation in dense point clouds through local feature extraction and point cloud density area analysis by means of a defect area clustering detection mechanism based on semantic guidance. According to the method, the accuracy and robustness of defect detection are remarkably improved, and the method is suitable for high-sensitivity automatic detection and three-dimensional positioning of surface defects of key parts of aerospace equipment.

The invention designs a method for predicting the health state of an energy storage battery, and the method comprises the steps: S1, carrying out the collection and monitoring of the charging and discharging process of an operation battery, and collecting and storing the charging and discharging capacity data of each cycle of the battery; s2, preprocessing the collected discharge capacity data of the batteries, calculating battery health state data, constructing a time sequence sample by using a sliding window mechanism, and dividing the battery health state data of each battery into an input sequence and an output sequence; s3, constructing a hierarchical neural network architecture and training the hierarchical neural network architecture; s4, performing Bayesian optimization on the hierarchical neural network architecture based on an Optuna framework; and S5, predicting the energy storage battery by using the optimized neural network architecture to obtain the health state of the battery. Through the above design, the prediction precision of the battery health state can be improved, and the stability and robustness of overall prediction are enhanced.

The invention discloses a heterogeneous processing unit dynamic scheduling method and system of a hierarchical neural network, and belongs to the technical field of heterogeneous computing resource scheduling. The invention aims to solve the problems of non-uniform task allocation, low resource utilization rate and poor real-time performance of a deep neural network in a heterogeneous computing environment. A static-dynamic scheduling strategy in which neural network task hierarchical division, an execution time prediction model based on machine learning, scheduling optimization objective function construction and a genetic algorithm and simulated annealing algorithm are combined is mainly adopted. According to the method, efficient dynamic scheduling of the neural network reasoning tasks among heterogeneous processing units such as a CPU, a GPU, an NPU and an FPGA can be achieved, and the overall calculation efficiency and the resource utilization rate are improved.

The embodiment of the present application provides a method and device for predicting the ablation wear amount of a gun barrel based on a combined model. By combining an adaptive neuron growth mechanism and a dynamic programming algorithm, a double-layer prediction architecture is constructed. The ablation wear amount is identified through a hierarchical neural network trained by temperature, pressure, and charge characteristic parameters, and an energy state constraint function is constructed based on sensor data. The system uses an escape search optimization operator to select the optimal wear evolution path, and establishes a multi-factor early warning threshold model by combining wear rate and acceleration indicators, realizing accurate prediction and evaluation and early warning of the ablation wear state. This method effectively solves the deficiencies of traditional technologies in model integration and state early warning, and provides a reliable basis for gun maintenance decision-making.

A method of training a hierarchical neural network in hardware comprising a plurality of neurons having tree structure connections between respective ones of the plurality of neurons to compute plausible inferences based on input data. The method includes initiating a WTA neuron ensemble to connect to data input neurons and produce a winner-take-all, attaching data records match with top neurons and branching out at WTA junctions to form a hierarchy network tree, converting a stream of the input data into neuron bipolar signals, updating neuron weights with Hebbian and anti-Hebbian rules, and repeating the attaching, converting, and updating until the input data is exhausted.

The application relates to the field of geomagnetic vector measurement, and discloses a magnetic target intelligent positioning method and system based on a layered neural network architecture. The method comprises the following steps: acquiring multi-point magnetic field data measured by a magnetic sensor array; calculating magnetic gradient tensor data according to the magnetic field data and the spatial position of the sensor; inputting the magnetic gradient tensor data into a pre-constructed layered neural network model; extracting local space-time features from the magnetic gradient tensor time series data of each measurement point through a sub-network layer; performing time series fusion and global context modeling on the feature sequences output by all sub-network branches through a global fusion network layer; and outputting the three-dimensional spatial position coordinates of the magnetic target according to the fused global features through an output layer. The implementation of the application is a deep integration of physical analysis methods and data-driven methods, can guide the mapping relationship between the magnetic gradient tensor and the position of the magnetic target of the deep learning model, and retains the strong nonlinear fitting capability, thereby enhancing the positioning accuracy of the magnetic target.

The invention relates to the technical field of aquaculture, in particular to an aquatic product gradient temperature control system and method for deep feature extraction of heat pump data, and the system comprises a data collection module, a feature extraction module, a deep learning prediction module, an energy efficiency optimization module, a fuzzy control module and an execution control module. By collecting heat source side inlet and outlet temperature, user side inlet and outlet temperature, system flow and system power data, the data are preprocessed, and multi-dimensional feature vectors including temperature gradient features, temperature difference features and frequency domain features are extracted. And predicting the energy efficiency ratio of the system by using a hierarchical neural network structure, constructing a temperature difference-energy efficiency mapping relation, and determining an optimal temperature difference set value. Control parameters are generated through the three-dimensional fuzzy inference engine, the rotating speed of the water pump and the running state of the heat pump are adjusted, gradient temperature control of the aquaculture environment is achieved, and the aquaculture requirements of high-value aquatic product species sensitive to temperature can be met.

The invention relates to a self-adaptive agile control method for a high-speed maneuvering flight-oriented intelligent unmanned aerial vehicle with a body, and the method comprises the steps: building a four-rotor unmanned aerial vehicle kinetic model, and outputting the coupling characteristics of aerodynamic force and torque and the prediction information of a future motion state; constructing a reinforcement learning strategy of the Lyapunov stability constraint, and obtaining a compensation control strategy according to output information and tracking errors of the kinetic model; constructing a hierarchical neural network architecture, and generating a motion control instruction of the quad-rotor unmanned aerial vehicle based on a compensation control strategy and latest state information fed back by a sensor; non-linear interference in flight is estimated through a non-linear interference observer, the non-linear interference is fed back to an adaptive control law in the hierarchical neural network, and control input of the hierarchical neural network is optimized in a prediction window through a rolling horizon optimization algorithm; and generating a motion control instruction for intercepting the target unmanned aerial vehicle by using the compensated and optimized adaptive control law and the control input, and sending the motion control instruction to an execution mechanism of the four-rotor unmanned aerial vehicle.

A hierarchical neural network for predicting out of stock products comprises an input layer that receives data from data sources that store disparate datasets having different levels of attribute detail pertaining to products for sale in stores of a retailer. A first level of neural networks processes the data from the data sources into respective learned intermediate vector representations. A second level comprises a concatenate layer that concatenates the learned intermediate vector representations from the second level into a combined vector representation. A third level comprises a feed forward network that receives the combined vector representation and outputs to the retailer an out of stock probability indicating which store and product combinations are likely to have out of stock products over a predetermined timeframe.

This paper discloses a method for constructing a lattice thermal conductivity model based on composite elemental and structural descriptors. The method comprises: obtaining elemental and structural information of a material; analyzing the material's elemental and structural information to extract composite descriptors; screening the composite descriptors using a genetic algorithm to determine important descriptors and other descriptors; generating an ensemble descriptor set based on the important descriptors and other descriptors; and constructing a lattice thermal conductivity model by hierarchically learning the ensemble descriptor set using a hierarchical neural network. This method constructs a highly accurate statistical ensemble hierarchical neural network that can accurately and efficiently predict lattice thermal conductivity based on elemental composition and structural information, thereby accelerating the discovery and development of new materials.

This invention discloses an intelligent analysis and control system and method for the entire lifecycle of gas wells based on artificial intelligence, relating to the field of oil and gas field development technology. It constructs a differentiable physical enhancement model, in which the gas well physical equations are embedded as differentiable hierarchical neural networks. Using historical data for training, a digital twin model is constructed that conforms to physical laws and possesses data fitting capabilities, and the digital twin model is end-to-end differentiable. Based on the differentiable physical enhancement model, a safety rating function is designed, mapping each state-action pair to a continuous safety score. This invention, by constructing a differentiable physical enhancement model, integrates gas well physical laws with data-driven capabilities, ensuring that state prediction and strategy output conform to physical mechanisms, avoiding unauthorized operations, and improving control reliability. Through the safety rating function and safety rating strategy gradient algorithm, safety constraints are implemented in the strategy optimization process, mitigating the impact of unsafe actions and ensuring gas well production safety.

The invention provides a medical image processing device, a hierarchical neural network, a medical image processing method, and a program, which can realize high-precision real-time processing of region of interest detection and category classification. The medical image processing device performs the following processing: acquiring a medical image; the method comprises the following steps: processing a medical image by using a feature extraction network of a hierarchical neural network, and extracting a first feature quantity and a second feature quantity with a resolution relatively higher than that of the first feature quantity from the medical image; detecting a region of interest included in the medical image by processing the first feature amount using a first sub-network of the hierarchical neural network; and classifying the region of interest by processing the second feature amount using a second sub-network of the hierarchical neural network.

The present invention provides a method for detecting the blurriness of a face. Based on adding blur segments and a hierarchical neural network training method, the aligned face grayscale image, LBP image, and Sobel feature map are used as inputs of a preset neural network model, the face blur category is used as the output of a first neural network, and the face blurriness is used as the output of a second neural network for model training. When the loss function of the preset neural network model tends to be stable, the model training is terminated to obtain the required face blurriness detection model. The present invention can effectively expand the distance between blur segments, improve the accuracy of face blur detection, and especially greatly improve the accuracy of face motion blur detection. It can be convenient to screen out images that meet the needs of face recognition, thereby improving the accuracy of face recognition, and has good practical value.

Systems and methods for processing classification field values in machine learning applications, particularly neural networks, are disclosed. Classification field values are typically converted to vectors prior to being passed to the neural network. However, low-dimensional vectors limit the ability of the network to understand correlations between contextually, semantically, or featured similar values. On the contrary, the high-dimensional vector may suppress the neural network, resulting in the network finding a correlation with respect to individual dimension values, which may be false. The disclosure relates to a hierarchical neural network comprising a primary network and one or more secondary networks. Classification field values are processed in the secondary network to reduce the dimensions of the values prior to being processed by the primary network. This enables contextual, semantic and feature dependencies to be identified without overloading the entire network.