23 results about "Dynamic neural network" patented technology

Apparatuses, systems, and techniques to assign a processing resource to an inference request directed to a neural network based on an amount of information to be inferenced indicated by said request. In at least one embodiment, an AI application is deployed with a software wrapper that intercepts inference requests and dynamically distributes such requests among available processing resources such as host processor(s) and / or AI accelerator(s) to improve execution performance of said AI application.

The invention provides a heterogeneous perception adaptive low-bit neural network deployment method, and relates to the technical field of artificial intelligence and heterogeneous computing, and the method comprises the steps: firstly obtaining the hierarchical computing feature information of each layer of a neural network and the dynamic feature parameter information of heterogeneous hardware, and forming a multi-level basic information set; and then a hierarchical efficiency association model is constructed to describe the association relationship among the calculation precision, the hardware dynamic characteristics and the layer calculation efficiency. During online operation, a hardware real-time load state and an energy efficiency constraint condition are tracked to generate a dynamic state monitoring result, and the dynamic state monitoring result and the dynamic state monitoring result are combined to generate a hierarchical deployment configuration scheme by adopting a multi-objective optimization algorithm. And finally, a lightweight runtime scheduler is called to allocate calculation tasks according to the scheme, and interlayer dependency data is loaded and executed, so that dynamic neural network deployment across hardware equipment is realized, and the deployment effect and the operation efficiency are improved.

The application provides a robot driving control method, device, equipment and storage medium, comprising: calculating a tracking error according to a desired trajectory and an actual running trajectory corresponding to each joint of a current time of a linkage robot; calculating the tracking error by using a sliding mode control algorithm to obtain a first input torque of the current time; taking an actual input torque of a previous time and the actual running trajectory as inputs of a dynamic neural network to predict an unknown torque in a robot dynamics model at the current time, and taking the predicted unknown torque as a second input torque of the current time; determining an actual input torque of the current time according to the first input torque and the second input torque, and controlling the linkage robot to move according to the actual input torque of the current time; and solving the problem of stability and accuracy decline of the linkage robot caused by various uncertain factors in the prior art.

The application provides a heterogeneous-aware adaptive low-bit neural network deployment method, and relates to the fields of artificial intelligence and heterogeneous computing technology. First, hierarchical computing feature information of each layer of a neural network and dynamic characteristic parameter information of heterogeneous hardware are acquired to form a multi-level basic information set. Then, a hierarchical performance correlation model is constructed to describe the correlation between computing accuracy, hardware dynamic characteristics and layer computing performance. During online operation, real-time hardware load state and energy efficiency constraint conditions are tracked to generate dynamic state monitoring results. A multi-objective optimization algorithm is used in combination with the two to generate a hierarchical deployment configuration scheme. Finally, a lightweight runtime scheduler is called to allocate computing tasks, load inter-layer dependent data and execute according to the scheme, thereby realizing dynamic neural network deployment across hardware devices and improving deployment effect and operation efficiency.

The invention provides a gas turbine state intelligent early warning method and system based on a dynamic neural network. The method comprises the steps of gas turbine historical operation data acquisition, characteristic parameter screening, neural network prediction model construction and training, gas turbine real-time data acquisition, gas turbine key operation state characteristic parameter prediction and analysis, dynamic updating of a gas turbine historical database and a neural network prediction model and the like. And real-time monitoring and early warning of the running state of the gas turbine are realized. According to the method, through boundary condition-operation state characteristic parameter classification and neural network prediction model dynamic modeling and training, the problems of limitation of single-parameter early warning in traditional gas turbine state monitoring, lack of prediction model updating mechanisms and the like are solved, and a technical means can be provided for intelligent operation and maintenance of the gas turbine.

The invention discloses a coagulant self-adaptive adding control method and system based on a physical constraint neural network, and relates to the technical field of water treatment automatic control, and the method comprises the steps: collecting water quality data; preprocessing the water quality data to obtain a standardized water quality sequence; inputting the standardized water quality sequence into a pre-trained time sequence dynamic neural network model, and predicting future dosage and future effluent turbidity; at the end of each control period, based on a model prediction control mechanism, performing rolling optimization on the pre-trained time sequence dynamic neural network model by taking the optimization objectives that the future effluent turbidity predicted by the time sequence dynamic neural network model does not exceed the standard and the future dosing amount is minimum, and solving an optimal medicament dosing sequence; and converting the optimal medicament feeding sequence into a control signal to drive a dosing pump to feed medicaments. The problems that a traditional control method is weak in self-adaptive capacity, poor in predictability and high in medicine consumption when facing complex and changeable raw water quality are solved, and stable standard reaching of the effluent quality and accurate optimization of the coagulant adding amount are achieved.

The application discloses a motor driving sudden change friction characteristic modeling and feedforward compensation method, which realizes decoupling and independent description of non-sudden change characteristics and sudden change characteristics of friction, and constitutes a dynamic neural network friction model with a width neural network structure which integrates the width neural network structure characteristics and the dynamic recurrent neural network structure characteristics; and the motor control system feedforward control compensation based on the friction model: the motor rotation angle is used as the input signal of the friction model, the friction model obtains corresponding friction torque prediction and estimation, the output of the motor controller is corrected through the feedforward control compensation, and the influence of internal friction on the motor execution precision is offset.

The application discloses a kind of bionic dynamic neural network, the neural network can form behavior state relation net by learning, and can be received when setting target task, state transfer and behavior activity control are carried out according to current state and the behavior state relation net formed, directly complete the target task.The bionic dynamic neural network provided in the embodiment of the application imitates the neural circuit topological structure and dynamic characteristics of biology, realizes the control to state transfer, so that state transfer and action control form the behavior state relation net with causal relationship, more in line with the way of human brain processing problem, improve the decomposition ability of complex task and the efficiency of reinforcement learning, so that reinforcement learning can be more realistically used in actual robot task.The bionic dynamic neural network of the embodiment of the application can be applied in reinforcement learning model, for building the reinforcement learning algorithm model with higher learning efficiency capable of decomposing task.

Apparatuses, systems, and techniques to assign a processing resource to an inference request directed to a neural network based on an amount of information to be inferenced indicated by said request. In at least one embodiment, an AI application is deployed with a software wrapper that intercepts inference requests and dynamically distributes such requests among available processing resources such as host processor(s) and / or AI accelerator(s) to improve execution performance of said AI application.

The invention provides a real-time speech enhancement method and system based on an adaptive deep neural network, and belongs to the technical field of speech signal processing. The method comprises the following steps: collecting multiple paths of voice signals through a multi-microphone array, and suppressing environmental noise by using a beam forming algorithm; processing the voice signal by adopting a hybrid noise reduction method of wavelet transform and adaptive filtering; constructing a dynamic neural network model, dynamically adjusting the network weight according to the noise environment classification result, and enhancing the voice signal; frequency band energy compensation is carried out on the enhanced voice in combination with a psychological acoustic model, and auditory fatigue is reduced; and finally, reconstructing a time domain voice signal through inverse short-time Fourier transform, and carrying out post-processing to output a final enhanced voice. The method has the advantages of being high in adaptability, high in real-time performance, good in voice quality and the like, and is suitable for various scenes such as voice communication, voice recognition, intelligent voice assistants and the like.

The invention belongs to the field of fine control, and particularly discloses a soil water and fertilizer fine control system based on machine learning, which comprises a soil water and fertilizer data real-time acquisition module, a soil water and fertilizer content prediction module and a soil water and fertilizer fine control module. According to the method, the dynamic neural network model is constructed, and Bayesian regularization training, a back propagation technology and a gradient descent algorithm are combined to optimize the model, so that high-precision dynamic prediction of the water and fertilizer content of the soil is realized; pID (Proportion Integration Differentiation) control parameters are dynamically generated by adopting an elitism strategy genetic algorithm with a memory bank, a method for synchronously optimizing valve opening, harmonic suppression and water and fertilizer loss indexes by combining a multi-target genetic algorithm is combined, self-adaptive optimization and multi-target collaborative optimization of parameters of the water and fertilizer controller are realized, and dynamic adjustment is performed by combining real-time monitoring data. And the PID parameters are ensured to be always in an optimal state.

The invention provides a robot driving control method and device, equipment and a storage medium. The robot driving control method comprises the steps that tracking errors are calculated according to expected trajectories and actual moving trajectories corresponding to all joints of a connecting rod robot at the current moment; calculating the tracking error by adopting a sliding mode control algorithm to obtain a first input torque at the current moment; taking the actual input torque at the previous moment and the actual moving track as the input of a dynamic neural network, predicting the unknown torque in the robot dynamic model at the current moment, and taking the predicted unknown torque as the second input torque at the current moment; the actual input torque at the current moment is determined according to the first input torque and the second input torque, and the connecting rod robot is controlled to move according to the actual input torque at the current moment; the problem that in the prior art, stability and accuracy of a connecting rod robot are reduced due to various uncertain factors is solved.

This invention discloses a dynamic neural network event-triggered control method for motor load frequency under external disturbances and internal coupling, relating to the field of motor control technology. The method includes the following steps: obtaining the system matrix parameters of the motor system; establishing a motor state-space model to obtain discrete states; based on the discrete states, representing external disturbances and internal coupling in the motor system as a common disturbance term, and estimating the common disturbance term online using a neural network to obtain disturbance estimates; obtaining the trigger update time of the control signal based on the state deviation between the discrete states and their most recent trigger state; constructing a dynamic neural network state feedback controller at the trigger update time; substituting the dynamic neural network state feedback controller into the motor system model, establishing linear matrix inequalities, and implementing dynamic event-triggered control by solving the linear matrix inequalities, which can better cope with unknown disturbances and reduce communication burden.

The invention discloses a motor load frequency dynamic neural network event trigger control method under external disturbance and internal coupling, and relates to the technical field of motor control, and the method comprises the following steps: obtaining system matrix parameters of a motor system, building a motor state space model, and obtaining a discrete state; based on the discrete state, uniformly representing external disturbance and internal coupling in the motor system as a common disturbance term, and performing online estimation on the common disturbance term through a neural network to obtain a disturbance estimation value; based on the state deviation between the discrete state and the last trigger state, obtaining a trigger update moment of the control signal; constructing a dynamic neural network state feedback controller at a trigger updating moment; a dynamic neural network state feedback controller is substituted into a motor system model, a linear matrix inequality is established, dynamic event trigger control is implemented by solving the linear matrix inequality, and the control effects of better coping with unknown disturbance and reducing communication burden can be achieved.

The invention discloses a step-by-step field effect transistor modeling method and system based on neural network space mapping, and belongs to the technical field of microwave circuit and device modeling. Three types of core models are constructed in stages: a static neural network space mapping model and a dynamic neural network space mapping model are established as intermediate modeling; constructing a dynamic neural network model containing more hidden layer neurons as a final target model; and the direct current data, the small signal S parameter data and the large signal harmonic balance data are respectively adopted to carry out targeted training on the model in each stage, and finally a high-precision transistor large signal model is obtained. A constraint condition formula is introduced in the training process, and it is ensured that training in the subsequent stage does not change the training result achieved in the early stage. According to the method, one-stop modeling of the transistor large signal model of the field effect transistor is realized, the modeling efficiency and the model precision are effectively improved, and an efficient and reliable solution is provided for transistor modeling in microwave circuit design.

A method and system for detecting armed personnel and equipment based on a dynamic neural network is disclosed. The method includes the following steps: acquiring images of armed personnel at different distances using a high-resolution camera with variable zoom; labeling personnel and equipment in each image to form an equipment detection dataset; constructing a dynamic neural network model, which includes a first sub-network and a second sub-network, wherein the first sub-network is used to detect humans in the image; when the first sub-network detects humans in the image, extracting the human ROI and transmitting it to the second sub-network; the second sub-network is used to detect equipment using classifiers at different network depths according to different image resolutions; training the dynamic neural network model based on the equipment detection dataset to obtain a trained armed personnel and equipment detection model; and inputting the image to be detected into the armed personnel and equipment detection model to obtain the armed personnel and equipment detection result of the image to be detected.

The application discloses a kind of poultry hatching intelligent management system based on multi-modal data fusion, through environmental monitoring module acquisition temperature, humidity, CO2 concentration and so on multidimensional parameter, after sliding average filtering pretreatment, combine variety specificity environment parameter library, pass through time sequence dynamic neural network and construct multi-parameter coupling model, output hatching window span time;Sound recognition module uses multi-source noise adaptive suppression algorithm to extract calling signal, after time domain, frequency domain feature extraction, through CNN+LSTM model accurately identifies the hatching stage;Image processing module acquires infrared image, after denoising, enhancement pretreatment, utilize YOLO model detection chick, statistics hatching quantity and proportion;Hatching management module is based on the output of the first three, through reinforcement learning dynamic weight distribution algorithm, adapt to different stages of data value in whole cycle of incubation, intelligently identify earliest hatching poultry, accurately predict hatching time, optimize batch push row.

The application relates to the technical field of bridge engineering and discloses a bridge pile foundation numerical model construction method for uneven distribution of structural surfaces, which comprises the following steps: collecting and arranging a bridge pile foundation design database under uneven structural surfaces; classifying and dividing all data samples and carrying out pretreatment; establishing an NAR dynamic neural network structure, initializing neural network related parameters; adopting a Logistic chaotic mapping to initialize a beetle population; and obtaining a current updated optimal population position through subsequent processing. The bridge pile foundation numerical model construction method for uneven distribution of structural surfaces is combined with an NAR dynamic neural network structure, a chaotic mapping initialized population and a beetle algorithm, a calculation model most conforming to the bridge pile foundation design database under uneven structural surfaces is finally determined, the calculation model is high in accuracy and efficiency, and economic cost is saved.

This invention discloses a dynamic convolutional neural network (CNN) hysteresis modeling and compensation method for motor and reducer modules. First, a dynamic CNN hysteresis model is constructed, consisting of a concatenated convolutional hysteresis operator and a dynamic RBF neural network. Then, the moment-to-moment load torque of the motor and reducer modules is fed into the dynamic CNN hysteresis model to obtain the moment-to-moment predicted torsional angle. In step 3, the moment-to-moment predicted torsional angle is used to compensate for the moment-to-moment motor rotation angle of the motor and reducer modules, resulting in the moment-compensated motor rotation angle. This invention proposes a simple dynamic convolutional hysteresis operator and combines it with a dynamic RBF that can achieve strong nonlinear characteristic mapping to construct a hysteresis model for the motor and harmonic reducer modules. Through feedforward compensation control, the module hysteresis error is compensated, improving the accuracy of the joint modules.

The invention discloses a geological disaster risk assessment method and system based on multi-source data analysis, relates to the technical field of geological disaster prevention and control, quantifies the uncertainty of multi-source data according to an evidence theory, updates the credibility weight in a Bayesian manner, and adapts the data quality through an attention dynamic neural network. The method has the advantages that uncertainty characteristics of multi-source data are quantified through the evidence theory, the credibility weight of the data source is corrected in combination with a Bayesian dynamic updating mechanism, the reliability of the data source is improved, the reliability of the data source is improved, and the reliability of the data source is improved. Then, a dynamic neural network based on an attention mechanism is utilized to adaptively adjust a structure and a decision boundary according to data uncertainty, robust optimization is introduced to suppress low-credibility data error conduction, and meanwhile key influence factors are calibrated and marked through credibility intervals; the defects that in the prior art, data uncertainty is ignored, an evaluation model cannot adapt to data quality fluctuation, only a single result is output, and no reliable basis exists are effectively overcome.

The application relates to a multi-view expression recognition method based on a dynamic neural network, which comprises the following steps: acquiring an image to be detected; performing face detection on the image, segmenting a face image and performing face region alignment; performing image preprocessing on the aligned face image; constructing an expression recognition model based on a deep convolutional neural network and performing training; the expression recognition model comprises a shallow feature extraction layer, a plurality of blocks composed of a convolution sampling module and a residual unit and a full connection layer which are connected in sequence, the preprocessed image is taken as input, and an expression recognition result is output, wherein the convolution sampling module dynamically generates a mask based on an input feature map, the generated mask is used for controlling the convolution position of a convolution layer in a corresponding residual unit, and dynamic range convolution is realized; and the trained expression recognition model is used for multi-view expression recognition. Compared with the prior art, the application has the advantages of high recognition accuracy, good robustness and the like.

The application relates to the technical field of rigid body modeling analysis, and discloses a rigid body modeling analysis system based on agricultural transmission shaft processing, which comprises the following steps: dividing a transmission shaft into M subsegments along a central axis based on segmentation constraints, and obtaining structure data of each subsegment, including an outer diameter, an inner diameter, a length and a material density; taking a central origin of the central axis of the transmission shaft as an X axis, taking two transverse directions perpendicular to the central axis as Y and Z axes, and taking the central axis as a coordinate system; combining and processing X-axis rotational inertia, Y-axis rotational inertia and Z-axis rotational inertia to form a rotational inertia tensor of the transmission shaft; inputting the rotational inertia tensor, an external load torque sequence and a load torque sequence of the transmission shaft under a target working condition into a pre-trained dynamics neural network model to obtain a predicted angular velocity sequence and a predicted rotation angle sequence; and performing check processing on the maximum shear stress of the transmission shaft based on the predicted angular velocity sequence and the predicted rotation angle sequence.

The application belongs to the field of computer vision, and particularly relates to a kind of based on dynamic neural network extension's continuous small sample image classification method and system, aims at solving the problem that model overall is difficult to adjust when training new class in prior art.The application includes: training basic feature extraction network with sample with known class data label, feature extraction is carried out to sample with known class data label by feature extractor, then projection layer is projected into expandable feature space and unused area is kept;Sample with new class data label and sample with known class data label are combined for training;New class visual feature is projected into unused area by projection layer of extended part parameter and aligned with known class, and classification is carried out by incremental class prototype classifier.The application is simple and flexible, can significantly improve the classification performance of new class object, and can effectively improve the classification performance of historical class object.