207 results about "Model switching" patented technology

The invention discloses a method for intelligently regulating and controlling production parameters in a fruit concentrated juice production process, which comprises the following steps of: acquiring multi-dimensional process parameters such as temperature, pressure, flow, concentration, equipment state and the like in real time through a multi-channel sensor network, and forming a standardized data sequence after filtering, normalization and drift correction; extracting stage features by using technologies such as a sliding window and Fourier transform, and inputting the stage features into the lightweight classification model to realize production stage identification; in combination with an identification result, dynamically calling a corresponding multi-target optimization sub-model, and realizing nonlinear prediction and optimal solution selection of process parameter setting by adopting an LSTM and a multi-target genetic algorithm; on the basis of real-time feedback, the performance of the model is automatically evaluated, self-adaptive adjustment and optimization of the optimization algorithm are achieved through reinforcement learning and an incremental updating mechanism, multi-target collaborative optimization, self-adaptive adjustment and model switching in the production process can be achieved, and the consistency of production efficiency and product quality is improved.

The invention discloses a method for supporting fast switching of multiple models by NPU (Network Processing Unit) firmware. According to the invention, the system real-time performance and the resource utilization rate in a multi-task scene are obviously improved. Through a dynamic hierarchical caching strategy and a priority scheduling mechanism, the system can intelligently allocate cache resources and preferentially guarantee rapid loading of a high-frequency and high-urgency model, for example, in an automatic driving scene, switching delay of a path planning model can be reduced to a millisecond level, and non-perceptual switching of key tasks is guaranteed. The incremental parameter loading technology and firmware-level context management are deeply fused, only model difference data are transmitted, and hardware is utilized to accelerate and recover a calculation state, so that bandwidth waste caused by traditional full-amount loading is greatly reduced, the model switching efficiency of edge computing equipment is improved by more than 10 times when multiple tasks such as voice recognition and image processing are carried out in parallel, and the efficiency of the edge computing equipment is improved. And meanwhile, the calculation precision is kept lossless.

Improved solutions that enable more effective and efficient communications with users, in particular with respect to field diagnostics and services. Some solutions can enable users to better communicate with a provider to obtain more useful diagnostic and service information. Certain solutions can employ multi-model switching machine learning techniques to enhance a user's communication with the provider and / or the provider's response.

The invention discloses a water surface target tracking method and system based on multi-physical parameter fusion, and the method comprises the steps: obtaining a physical layer observation parameter of a target through a water surface monitoring radar, and constructing a multi-dimensional nonlinear observation vector; establishing a multi-mode motion model set including constant speed, variable speed and turning, and realizing dynamic switching among motion modes through a Markov chain; adjusting a process noise covariance and an observation noise covariance based on a residual covariance estimation result in the sliding window by adopting an adaptive extended Kalman filtering algorithm; distributing weights according to the measurement variance of each physical parameter, optimizing the Kalman gain through a weighted least square method, and completing the updating and estimation of a target state; dynamic switching of motion modes is achieved through a Markov chain, and when state estimation residual errors of continuous preset times exceed a preset threshold value, model mismatch is judged, and a Markov chain model switching mechanism is triggered; according to the method, the sea condition adaptability, the calculation efficiency and the engineering expandability can be improved.

The invention discloses a cable sheath ring current fault positioning in-loop simulation method and system based on LABVIEW. The method comprises the following steps: collecting ring current data of cable sheath monitoring points in real time under the control of LABVIEW, and carrying out digital display and graphical display; decomposing circulation data to obtain a wavelet coefficient and calculating features, and inputting a classification model to obtain a fault type when the features are abnormal; calculating a compensation coefficient considering the influence of the electromagnetic field, and calculating the position of a fault point in combination with a double-end traveling wave method; the operation model switches a typical working condition simulation mode, prediction of a Bayesian reasoning correction model is carried out, and the full-life-cycle operation state of the cable is simulated; and performing fault early warning through the LSTM model and risk assessment. The method can effectively improve the cable operation and maintenance efficiency and reliability, and is widely applied to cable fault detection and guarantee of safe and stable operation of a power system.

The present disclosure describes methods of data-driven AI / ML model signaling for efficient model switching with machine learning model operation in wireless mobile communication systems including base stations (e.g., gNB) and mobile stations (e.g., UE). The pre-configured model modes with the associated candidate models are used to switch models based on model monitoring.

The invention belongs to the technical field of intelligent unmanned systems and edge intelligent reasoning, and discloses an unmanned aerial vehicle emergency inspection system integrating low-altitude communication and edge calculation. The system is composed of a link state monitoring and scoring module, a reasoning calculation migration and model loading module, a task graph modeling and dynamic segmentation module, a multi-unmanned aerial vehicle task collaboration and trajectory continuation module, a flight state fusion and path adjustment module, a redundancy criterion driven fault-tolerant control module and a result consistency verification and data caching module. According to the method, the communication link scoring function fusing the packet loss rate, the bandwidth, the time delay and the signal-to-noise ratio is constructed, the comprehensive quality evaluation of different low-altitude communication links is realized, and the reasoning model switching and task migration strategy is driven by the scoring result, so that the communication adaptation capability of the system in a complex, dynamic or emergent scene is remarkably improved, and the system performance is improved. Compared with an existing communication mechanism depending on a fixed link or a preset priority.

The invention provides an AI reasoning optimization method and system for a dynamic model switching framework for edge equipment, and relates to the technical field of AI optimization, and the method comprises the steps: training a quantization compensation model through federated learning to correct a quantization error; constructing a hierarchical heterogeneous resource management system to realize multi-core collaborative scheduling; the compensation parameters are integrated to an FPGA acceleration circuit, and an intelligent power consumption equalization technology is adopted; and constructing a dynamic decision engine by using the graph neural network to trigger model switching. According to the invention, the balance between high-precision AI reasoning and low-power consumption requirements on the edge equipment is realized, the utilization efficiency of computing resources is improved, and the energy consumption is reduced.

The invention discloses a multi-energy access electric power spot market risk prevention and control method and system, and relates to the technical field of risk prevention and control, and the method comprises the following steps: obtaining multi-dimensional data of a to-be-detected region, obtaining the risk conduction intensity between nodes based on a power grid topological structure, and constructing a dynamic risk topological network model; according to the risk topology network model, node operation characteristics are extracted, evaluation data of the classification model are synchronously obtained, model switching regulation and control are carried out according to the evaluation data, and the model switching regulation and control further comprises the step of outputting a risk overflow probability; through the method based on the dynamic risk topology network and multi-model adaptive switching, the problems of inaccurate risk prediction and untimely response in multi-energy access to the electric power spot market are solved.

The invention discloses a multifunctional edge computing device and method based on artificial intelligence, and particularly relates to the technical field of wind power operation management, and the method comprises the following steps: deploying a main model and an auxiliary model at an edge node, enabling the main model to execute real-time reasoning, and enabling the auxiliary model to be used for credibility evaluation; setting a disturbance detection mechanism, injecting a disturbance sample with preset characteristics into the main model, comparing the disturbance sample with the output of the auxiliary model, and calculating a response offset degree; fusing the offset degree and the reasoning difference to generate a credibility score; when the score is lower than a threshold value, automatically selecting to execute parameter reloading or switch an auxiliary model, and reconstructing a reasoning session to guarantee stable operation of edge nodes; according to the method, real-time perception and structural offset judgment of the operation state of the edge computing device are achieved through disturbance detection, main and auxiliary model collaborative verification and a credibility scoring mechanism, parameter reloading or model switching is dynamically executed based on a strategy tree, and the stability, self-diagnosis capability and fault tolerance level of the model in an extreme environment are remarkably improved.

The invention relates to the technical field of data processing, and discloses an automobile industrial database rapid management method and system based on virtual-real mapping. The method comprises the steps that a vehicle type mixed line recognizer recognizes the vehicle type switching state of a production line, and dynamic data of a vehicle type identifier, a process path and a resource demand are acquired; adjusting a partition strategy of the heterogeneous database according to the dynamic data, and generating a partition mapping table and an access weight matrix; binding the mapping table with the three-dimensional twin model through a time sequence association algorithm to form an automatically switched virtual-real mapping object pool; predicting a cross-vehicle-type data demand based on the object pool, and generating a preloading queue and a conflict avoidance strategy; and the vehicle type switching trigger processes the preloading queue to realize zero-delay data switching and consistency verification. The response speed and the data access efficiency of database management in the vehicle model line changing process are improved.

The invention relates to an unmanned ship cluster cooperative tracking control method based on model predictive control and control law learning, and belongs to the field of unmanned ship formation trajectory tracking control. The method aims at solving the problems that an existing centralized control algorithm is high in calculation complexity, and a distributed control algorithm is poor in control effect. Firstly, kinematics and dynamics models of the unmanned ship are established, then a distributed dual-mode control algorithm and a centralized model predictive control algorithm are designed, a long short-term memory (LSTM) network is constructed to learn a deep mapping relation between a global control law and a distributed control law, and then, the deep mapping relation between the global control law and the distributed control law is learned by aiming at the problem that an offline learning method is poor in expandability. A self-adaptive model switching strategy is designed, the LSTM model is switched in a self-adaptive mode according to the characteristics of the tracking trajectory, the cooperative control effect is improved while the calculation time is shortened, and it is ensured that the multiple unmanned ship systems can track the time-varying target trajectory in real time according to various expected formations.

The invention discloses a cooperative fault-tolerant positioning method based on self-adaptive residual model interaction, and the method comprises the steps: enabling single fusion information of each sensor to form a sub-model based on a model mixing thought; residual interaction is carried out on each sub-model by using an interactive multi-model, and the model of the fault sensor is switched, so that a relatively large positioning error is prevented from being introduced in a cooperative positioning process; and finally, obtaining optimal state estimation of cooperative positioning according to model state probability fusion. In order to improve the model switching capability, Markov transition probability estimation of adaptive parameter adjustment is introduced into an interactive multi-model framework, so that the situation that the cooperative fault-tolerant performance is reduced due to model switching failure is avoided. The method can be used for cooperative fault-tolerant positioning of the unmanned aerial vehicle cluster in a limited satellite and unknown complex and easily-interfered environment.

Provided in the embodiments of the present specification are a task processing method, an automatic question answering method, and an image generation method. The task processing method, which is applied to a control unit in a task processing platform, comprises: in response to a task processing request, determining a plurality of task processing models corresponding to the task processing request; analyzing the plurality of task processing models to determine the current task processing model; acquiring target model data of the current task processing model from a storage unit, and copying the target model data to a processing unit, such that the processing unit generates a processing result of the current task processing model on the basis of the target model data; and when the processing unit satisfies a preset model switching condition, releasing the target model data from the processing unit, and returning to execute the step of analyzing the plurality of task processing models to determine the current task processing model until a preset stop condition is reached, so as to obtain a task processing result corresponding to the task processing request. The complexity of a task processing process is reduced.

Along with diversification and complexity of underwater target motion forms, an existing interactive multi-model algorithm has the problems of slow model switching and insufficient tracking precision when facing target state switching. Therefore, the invention provides a parallel interactive multi-model target tracking algorithm based on time sequence information by taking a classical IMM algorithm as a main body framework. According to the algorithm, model probability change trends at adjacent moments are compared, parameters of a state transition matrix are dynamically corrected, and self-adaptive updating of the state transition matrix is achieved through normalization processing. And meanwhile, the model probability is dynamically updated by utilizing a parallel IMM framework and information entropy, so that the tracking precision reduction caused by excessive correction of a state transition matrix is avoided. Simulation results show that compared with an existing algorithm, the algorithm provided by the invention has the advantage that the prediction precision of the target is improved by 3.52% to 7.87%. And meanwhile, the switching speed of the model is higher, and the underwater target tracking precision is effectively improved.

A method for a first apparatus, the method comprising: obtaining (713) at least one matching factor, wherein the at least one matching factor indicates a degree matching between inputs used during a machine learning training phase and inputs used during a machine learning inference phase; obtaining at least one threshold matching factor from a second apparatus; comparing (715) the at least one threshold matching factor and the at least one matching factor; performing at least a machine learning model switching based on the monitoring decision request from the second apparatus, the monitoring decision request based at least on: the obtained at least one matching factor; and at least one channel characteristic.

The invention provides a cross-scale digital twinning mixed operation method and system for a network construction type wind turbine generator, and relates to the technical field of wind turbine generators, a network construction type wind turbine generator digital twinning model comprising a refined converter model and an average value converter model is built, and a direct current bus current is introduced to construct a dynamic threshold triggering mechanism; when the instantaneous change rate exceeds a dynamic threshold value, the converter is triggered to be switched from an average value model to a refined model, and state synchronous compensation is carried out; and when the instantaneous change rate is lower than a dynamic threshold value and exceeds a preset duration, triggering switching from the refined model to the average value model, and performing state average value reinjection. Therefore, the hybrid modeling method based on multiple time scales realizes model switching through a dynamic threshold triggering mechanism, ensures data continuity and smooth transition based on a state synchronous compensation and state mean value reinjection strategy, improves calculation efficiency, and reduces tracking errors.

The invention discloses a permanent magnet motor direct torque control method and system based on model prediction and a medium, and belongs to the technical field of permanent magnet synchronous motor control. According to the permanent magnet synchronous motor direct torque control method based on model prediction, two-phase current of a permanent magnet synchronous motor is processed by constructing a Clarke transformation model, a flux linkage and torque model, a voltage vector optimization model, a switching time calculation model, a cost function minimization model and a three-phase inverter control model, a space voltage vector is determined, and the direct torque of the permanent magnet synchronous motor is obtained. And switching action information of a switching tube is obtained and is used for controlling bus voltage to be input into the permanent magnet synchronous motor, so that the permanent magnet synchronous motor is driven, and direct torque control of the permanent magnet synchronous motor based on model prediction is completed, and therefore, the switching state of a switch can be controlled to be optimal, and torque pulsation is effectively reduced; therefore, the system operation efficiency is effectively improved, the system stability is improved, the switching loss is reduced, and the scheme is scientific, reasonable and feasible.

The embodiment of the invention provides a model weight updating method and device, a storage medium and a program product. In the embodiment of the invention, the first original model is compiled and optimized in advance to obtain the target model without weight information, and the second original model isomorphic with the first original model is started to execute the deep learning task, so that the second weight information corresponding to the second original model can be directly injected into the target model; the target model can execute the deep learning task on the basis of the injected second weight information, and for the isomorphic model, only one compiling optimization process needs to be executed, so that the purpose of free compiling during isomorphic model switching is achieved, and the compiling optimization time is saved; furthermore, the second weight information is injected into the target model to realize the switching of the isomorphic model while the performance benefit brought by compiling optimization is obtained, and the model deployment efficiency can also be improved.

The invention relates to a fresh air conditioner control system and method based on double-model safety evolution. The fresh air conditioner control system comprises a credible data acquisition module, a mechanism main model controller and a shadow AI model training and verification module. A dynamic anchoring control module; a man-machine collaborative decision-making interface; and a fault isolation and model switching module. On the premise of guaranteeing control safety, the invention provides an industrial air conditioner control system architecture with evolution capability, adaptive capability and deployment flexibility, breaks through the limitation of the traditional control technology in nonlinear, multivariable and strong coupling scenes, and has relatively high engineering application value and popularization potential.

The invention relates to the technical field of electric vehicle charging infrastructure intelligent management, and discloses a charging pile load balancing method and system based on a neural network, and the method comprises the steps: analyzing the state, environment and user behavior data of a charging pile through a multi-mode auto-encoder network; generating a unified scene representation; recognizing a typical scene by using a variational auto-encoder and a density clustering algorithm; decomposing the charging load balancing knowledge into cross-scene shared knowledge and scene specific knowledge; automatically generating a neural network architecture adaptive to the current scene based on the scene similarity; using a meta reinforcement learning algorithm to train a meta strategy network to generate a load balancing decision; scene smooth transition is realized through gradual model switching; according to the method, multi-scene adaptation, knowledge sharing and smooth switching are realized, the problems of system redundancy, knowledge isolation, scene switching performance fluctuation and the like in a traditional method are effectively solved, and the operation efficiency of the charging infrastructure is improved.

The invention relates to the technical field of cross of storage controllers and artificial intelligence acceleration, and discloses a large model reasoning system and method based on combination of a flash memory controller and an NPU (Network Processing Unit), and the large model reasoning system comprises the flash memory controller, the NPU and a flash memory array, the flash memory controller integrates a host interface module, a flash memory interface module, an independent AI acceleration interface module and an AI management engine, and the AI management engine autonomously completes NPU initialization, model weight direct loading, KV Cache hierarchical management, RAG knowledge base retrieval and model switching; the flash memory array is divided into a firmware partition, an AI special partition and a user storage partition, and different data storage requirements are met. According to the method, large model reasoning with low delay and low CPU dependence can be realized, and the model loading delay is reduced from 5-30 seconds to lt; after 500 milliseconds, the CPU occupancy rate of the host is reduced from 15-25% to lt; 2%, and concurrent operation of 4-8 models is supported. According to the invention, integration of storage and calculation is realized, and edge end, data center and mobile equipment scenes are adapted.

The invention relates to the technical field of FSRU regasification control systems, and particularly discloses an FSRU regasification heat source intelligent switching control system based on multi-source coupling optimization, and the system comprises a multi-source grading collection and edge preprocessing module which is used for collecting heat source-load-environment-heat storage-abnormity five-dimensional basic data and dynamic operation constraint data, and sending the data to a data processing module; carrying out hierarchical transmission and normalization processing on the data; and the dual-mode adaptive multi-source-heat storage coupling model module is used for pre-constructing a lightweight coupling model and a high-precision coupling model, and dynamically switching between the lightweight coupling model and the high-precision coupling model according to a working condition trigger signal. By introducing a transient asynchronous compensation mechanism, the system effectively solves the problem of control window period caused by time consumption of high-precision model calculation, ensures that the system can still maintain smooth and continuous instruction output at the moment of extreme load fluctuation, and eliminates the risk of pressure oscillation caused by model switching.

A smart home center multi-task processing method based on deep neural network fusion belongs to the crossing field of artificial intelligence and Internet of Things, and comprises the following steps: pruning an original model to generate pruning candidate models with different sparseness, and forming a model pool by the original model and the pruning candidate models; combining the pruning candidate models into a fusion model through weight sharing; training the fusion model in a multi-task joint training mode to recover the accuracy of the fusion model; integrating the fusion model completing multi-task joint training into a unified fusion model through re-fusion; and re-training the unified fusion model in a multi-task joint training mode to recover the reasoning precision of the unified fusion model. According to the method, a dynamic extensible unified fusion model is generated through deep neural network pruning, weight virtualization and refusion technologies, different hardware resource conditions can be adapted, and it is ensured that multiple tasks can be efficiently and accurately executed under the resource limited condition. According to the method, the model switching overhead is greatly reduced, and the task processing time delay is reduced.

The application discloses a low-altitude intelligent networking dynamic collaborative reasoning method based on multi-agent reinforcement learning, relates to the technical field of low-altitude intelligent networks and edge artificial intelligence, and comprises the following steps: preloading a light model and a complex model on each unmanned aerial vehicle (UAV), deploying a complete complex model by means of a ground station, and constructing an air-ground integrated intelligent reasoning system. By introducing an enhanced multi-agent deep reinforcement learning algorithm, each UAV can dynamically select a model type, determine a model segmentation point, and reasonably allocate bandwidth and ground computing resources based on the state of the UAV, network conditions and task characteristics during task execution, so that multi-DNN flow heterogeneous resource-aware collaborative reasoning is realized. The application can effectively improve reasoning accuracy and reduce average delay under different device performance, bandwidth conditions and task density, has good system scalability and adaptability, and solves the problems of limited single-machine processing capacity, restricted communication resources and inefficient model switching.

The invention discloses a fault prediction system and prediction method for a helium filling recovery system of helium mass spectrum equipment, and the system comprises a data collection module which is connected with a PLC of the helium mass spectrum equipment through an industrial communication protocol, and collects multi-source original data in real time; the data fusion processing module is connected with the data acquisition module and used for performing data cleaning, data segmentation, data association and data integration processing on the multi-source original data and performing intelligent classification on the original data based on a preset algorithm; and the fault prediction module is connected with the data fusion processing module, is used for carrying out intelligent analysis on the data, comprises a decision tree prediction unit, a convolutional neural network prediction unit and a model switching unit, and is used for outputting a fault probability or a fault type. The method not only can accurately predict potential faults, but also can easily adapt to different application requirements and technical upgrading through a flexible modular architecture, and can remarkably enhance the reliability and maintenance efficiency of helium mass spectrum equipment.

Provided in the embodiments of the present disclosure are a model weight updating method, a device, a storage medium and a program product. In the embodiments of the present disclosure, compilation optimization is performed on a first original model in advance to obtain a target model without weight information, and when a second original model isomorphic to the first original model is enabled to execute a deep learning task, second weight information corresponding to the second original model may be directly injected into the target model, so that the target model may execute the deep learning task on the basis of the injected second weight information. For isomorphic models, only one compilation optimization process is required to be executed, thereby achieving the aim of eliminating a requirement for compilation during switching between the isomorphic models and saving on the time for compilation optimization; furthermore, performance benefits caused by compilation optimization are obtained, and injecting second weight information into a target model to implement the switching between the isomorphic models can also improve the model deployment efficiency.

The embodiment of the invention provides a management operation execution method and device, a storage medium and an electronic device, and the method comprises the steps: determining a priority set for at least one model in a first node, and determining the occupation number of model processing units (MPU) of each model in the at least one model and the processing time of each model, the MPU occupation number and the processing time have a first association relationship; and executing a management operation on the at least one model through the priority and / or the first association relationship, wherein the management operation comprises at least one of the following: deactivating the model, adjusting the number of occupied MPUs, activating the model, selecting the model, switching the model, updating the model and adjusting the processing time. By adopting the technical scheme, the technical problem of how to manage a plurality of artificial intelligence models or computing units at the same time under the condition of limited computing, storage and other resources at the terminal side in the related technology is solved.

The utility model discloses an adjustable gasket for a slit type extrusion coating machine, which comprises a fixed gasket body provided with a first through hole; the left side gasket body is vertically arranged on one side of the fixed gasket body and is provided with a second through hole; the right side gasket body is vertically arranged on the other side of the fixed gasket body and is provided with a third through hole; the first adjustable gasket body is movably arranged above the gasket body on the left side; the second adjustable gasket body is movably arranged above the right side gasket body; the thickness of the fixed gasket body, the thickness of the left side gasket body, the thickness of the right side gasket body, the thickness of the first adjustable gasket body and the thickness of the second adjustable gasket body are equal. By means of the mode, the gasket does not need to be disassembled frequently every time when models are replaced, only the position of the adjustable gasket body needs to be changed, the model switching time is effectively shortened, the gasket does not need to be replaced integrally, the number of jigs needed for model replacement is not too large, and the design cost is effectively reduced.

The application relates to the technical field of low-orbit satellite communication, and discloses a satellite communication-oriented traffic prediction method, device, equipment and medium, which are applied to a satellite-borne system, wherein the method comprises the following steps: receiving a self-description model package sent by a ground terminal in a standby area, and keeping an old model currently running in an active area to execute an inference task; obtaining original traffic data in a preset sliding window, determining a data correction factor of an adaptive ground model based on the difference between the original traffic data and statistical information; loading the data correction factor to the ground model to obtain a corrected new model; and in response to a model switching instruction, updating a global active model pointer from a memory address pointing to the old model to a memory address pointing to the new model by atomic operation. The technical scheme provided by the application can realize seamless hot updating of a model from the ground to the satellite without interrupting real-time inference services of the satellite-borne system.