71 results about "Learning set" patented technology

The invention discloses an AI-based multimodal transport resource collaborative dynamic configuration method, which comprises the following steps of: constructing a real-time data layer for acquiring multi-dimensional data; preprocessing the data of the real-time data layer; dynamically constructing a digital twin platform based on the preprocessed data information; training a plurality of agents; carrying out cooperative training on a plurality of agents under a federated learning cluster framework; jointly training a global AI model; the AI decision center generates an optimal or nearly optimal dynamic resource configuration decision; the dynamic configuration engine dynamically schedules resources and generates an instruction; issuing the generated detailed instruction to a physical system of an execution layer; and the IoT equipment continuously monitors the execution state and the physical environment change, and feeds back new data to the real-time data layer. According to the invention, the bottleneck of data islands and response delay is broken through, and a cost-aging-carbon emission multi-target balanced intelligent decision-making system is constructed; and performing multi-agent collaborative training under a federated learning framework to realize cross-domain collaborative optimization.

The invention discloses an unmanned aerial vehicle cluster environment adaptive optimization method based on machine learning, which relates to the technical field of unmanned aerial vehicle cluster cooperative control and comprises the steps of basic framework construction, training parameter optimization, dynamic strategy adjustment, anti-interference communication enhancement and fault tolerance self-reconstruction. Coupling the optimized federated learning model with a multi-modal data fusion module, performing self-attention mechanism fusion on sensor data after time synchronization, constructing a high-dimensional state vector containing an agent state and an environment feature, inputting the high-dimensional state vector into a reinforcement learning framework to generate a joint action decision, and realizing fault detection through an LSTM network. And performing fault compensation by using redundant sensor data and the multi-modal fusion model. Through a dynamic graph attention mechanism and multi-target federated learning, the cluster can dynamically adjust a strategy to cope with complex environments such as electromagnetic interference and obstacle change, and a multi-agent collaborative decision and residual error compensation fault self-healing mechanism ensures that the cluster can still complete a task when a node fails or communication is interrupted.

The invention relates to a critical flight identification and guarantee plan inversion method based on LTOT, and the method comprises the steps: data integration and preprocessing, critical flight identification, multi-objective optimization, guarantee plan generation, real-time monitoring and feedback, emergency response and the like. Intelligent monitoring and optimal scheduling of flight states and airport resources are realized by using the multi-layer technology and large model layer technologies such as a deep neural network, reinforcement learning, ensemble learning and natural language processing, and the system has real-time data acquisition, risk prediction, dynamic threshold adjustment and a cross-department collaboration platform, can quickly respond to emergencies, optimizes resource allocation, and improves the scheduling efficiency. The flight punctuality rate and the operation efficiency are improved, the data security and privacy are ensured, and the intelligent level and the emergency response capability of aviation scheduling are remarkably improved.

The embodiment of the invention provides a federated learning contribution evaluation method and device, and the method comprises the steps: carrying out the grouping of a plurality of edge computing devices, and obtaining a plurality of sub-federated learning sets; and for the target federated learning sub-set, aggregating model update information corresponding to each edge computing device in the target federated learning sub-set, and determining a collaborative contribution value of the target federated learning sub-set based on the performance index of the updated global model on the common test set. Through a first linear programming solver, according to the collaborative contribution values of the multiple federated learning sub-sets, obtaining the maximum loss value corresponding to all the federated learning sub-sets and optimizing the maximum loss value to obtain the minimized maximum loss value, and through a second linear programming solver, obtaining the maximum loss value corresponding to all the federated learning sub-sets; and according to the maximum loss value after all the sub federated learning sets are minimized and the reference contribution values corresponding to the plurality of edge computing devices, target contribution vectors corresponding to the plurality of edge computing devices are determined, and the contribution degree of each edge computing device in the training process is accurately quantified.

The invention discloses a ground vegetation leaf area index remote sensing inversion method based on a remote sensing image. The method comprises the following steps: data acquisition and preprocessing; performing multi-scale space-time non-local filtering fusion; dynamic feature extraction; carrying out transfer learning fine tuning; time sequence deep learning integration; and model output and post-processing. According to the invention, through multi-source space-time fusion and dynamic feature distribution, vegetation LAI inversion with high resolution and high continuity is realized; the transfer learning and the time sequence deep network enhance the adaptability of the model to a new region and time sequence change; the stability and reliability of large-scale application are guaranteed through full-process automation and uncertainty evaluation, and the model is obviously superior to an existing single-source or static model.

The invention provides a feature reduction and generalized inverse technology-based phenotype prediction method and system for self-adaptive fusion of linear and nonlinear effects, and constructs a new model for self-adaptive fusion of linear and nonlinear effects of prediction factors for crop phenotype prediction. And the importance degree of different effects on the phenotype is evaluated through a self-adaptive weight adjustment method, so that the function of comprehensively considering the influence of the linear main effect and the nonlinear relationship in crop phenotype prediction is realized. When the method is used for solving the model, the training data set is subjected to feature reduction, and the model is rapidly and effectively solved by using the generalized inverse, so that the determination efficiency of the phenotype value of the biological material is improved. According to the invention, model learning is carried out through the learning set and the test set obtained by multiple times of random division of the experimental data set, the system error of the model is reduced, and the stability of the phenotype prediction result is improved. The effectiveness of phenotype prediction is verified on genotype data based on DNA molecular markers and intermediate omics data based on metabolites.

Aspects of the disclosure are directed to a tree-based network architecture for serving and / or training machine learning models. The architecture includes one or more multi-chip packages having a plurality of compute-memory stacks connected via an input / output (I / O) die. The I / O die includes an aggregator to aggregate computations from the compute-memory stacks. The architecture can further include a plurality of the multi-chip packages connected on a server via a server level aggregator and a plurality of the servers connected on a rack via a rack level aggregator for further aggregation of the computations from the compute-memory stacks. The tree-based network architecture allows for fewer hops, resulting in lower latency and savings in bandwidth when serving and / or training machine learning models.

The present invention relates to a hard disk fault prediction method and system based on multi-model fusion, the method comprising: obtaining hard disk SMART data of hard disks of different models; extracting features from hard disk SMART data based on progressive window sampling to obtain an initial feature set; selecting features from the initial feature set using an extreme random tree model to obtain an optimal feature set; training multiple machine learning models through the optimal feature set to obtain a hard disk fault prediction model based on multi-model fusion; when a target model hard disk needs to be predicted for a fault, obtaining the current SMART data of the target model hard disk and extracting the current feature set; inputting the current feature set into a hard disk fault prediction model based on multi-model fusion to output a fault prediction result of the target model hard disk. The feature extraction capability and time series modeling capability of the machine learning integrated model are fully utilized to capture the changing process of each SMART attribute of hard disks of different models and improve the accuracy of hard disk fault prediction.

The invention belongs to the computer image processing technology, and particularly relates to an intelligent image recognition and analysis system which comprises an image preprocessing module, a multi-scale feature extraction module, a feature fusion and screening module, an image classification and recognition module, an image analysis and understanding module, a real-time feedback and optimization module and the like. Multi-scale feature extraction and feature screening are respectively realized through an improved convolutional neural network and an attention mechanism, and a corresponding algorithm formula is given. And a deep learning classification algorithm, a graph neural network and a semantic analysis algorithm are adopted to improve the image classification and understanding ability. And in combination with technologies of data enhancement, transfer learning, integrated learning and the like, the system is optimized in real time by utilizing reinforcement learning, so that the image recognition accuracy, robustness and analysis depth can be effectively improved.

To provide a pile construction evaluation system estimating a pile construction result.SOLUTION: A pile construction evaluation system is provided with: a site information acquisition section acquiring construction place information and construction site test result information as site information; an estimation section, on the basis of dictionary information storing test result information including at least soil property per depth acquired through a test and an N value showing a result of a standard penetration test, information showing a test place, an occurrence result of a feature of a construction result upon pile construction at the site, pile position information showing positions of constructed piles, construction information on pile construction outputted with a construction machine having constructed piles, and constructed pile distances between positions where the standard penetration tests have been carried out and positions of constructed piles in association with each other, site information acquired with the site information acquisition section, and wight upon learning set according to the constructed pile distances, estimating a feature of a construction result occurring at the site; and an output section outputting a result estimated by the estimation section.SELECTED DRAWING: Figure 1

The invention discloses a hypergraph-based unknown network attack detection method for an electric power Internet of Things terminal, relates to the technical field of network attack detection, and solves the problems of insufficient model expression ability, overfitting and learning set deviation caused by the fact that a model method in the prior art ensures that known classes are fully separated and unknown classes are far away from the centers of the known classes. According to the method, dual modeling capabilities of a graph structure and a time sequence structure are combined, so that the method can effectively adapt to complex distribution characteristics in a dynamic electric power Internet of Things environment, and is particularly suitable for the problems of non-uniformity, burstiness, unknown traffic characteristic change and the like in an electric power Internet of Things terminal data stream; and the adaptability of the model to the diversified data structure of the edge device is improved. A dimension compression mechanism is introduced in the structural design of the model, the parameter quantity of the model is effectively reduced, lightweight deployment and end-side reasoning on an edge node or an industrial terminal are ensured, and therefore the real-time attack detection capacity of the power system is improved.

The invention belongs to the technical field of defect detection, and discloses a laser surface alloying molten pool defect detection method based on a physical model, and the process from molten pool forming and flowing to final solidification can be well reacted in the numerical simulation and computer simulation process. According to the simulation result, the coordinates of the solidified material can be obtained, and three-dimensional coordinates can be obtained by using the coordinates to analyze the morphology of the strengthened part and the powder contact part, so that the relevant physical characteristics of the edge defects of the molten pool are obtained. In order to improve the recognition accuracy of an existing molten pool infrared image, coordinates occupied by all components obtained through molten pool simulation can be combined with the features of the molten pool infrared image. The most common problems of less experimental data and insufficient defect feature diversity of a neural network method are solved, a learning set of the neural network is further expanded by combining a result obtained by computer simulation after a certain amount of experimental data exists, and the accuracy of molten pool defect identification is fundamentally improved.

The invention discloses a knowledge exchange method based on learning before forgetting and application thereof, and relates to the technical field of machine learning and knowledge management, a pre-trained deep learning model is selected as a basic model, and a retention set, a forgetting set and a learning set are constructed; the basic model carries out new knowledge learning on the learning set, and the training target of the model in the stage is that the accuracy of the learning set is close to 1, and meanwhile the accuracy of the reserved set is kept unchanged; after the basic model completes new knowledge learning, knowledge irrelevant to a new task is forgotten through a selective forgetting mechanism, and the training target of the model at the stage is that the accuracy rate of a forgetting set is close to 0, and meanwhile the accuracy rates of a reserved set and a learning set are kept unchanged. According to the method, continuous learning and machine forgetting are integrated together, so that the problem of contradiction between useless knowledge forgetting and new knowledge learning when a deep learning model processes knowledge updating of a pre-training model is solved. According to the knowledge exchange method based on first learning and then forgetting, the specified knowledge can be selectively forgotten while efficient learning of new knowledge is ensured, so that more refined knowledge regulation and control are realized, and the adaptability and the stability of the model are improved.

The invention discloses a digital intelligent engineer project practical training teaching method based on robot learning, and relates to the technical field of teaching. Task difficulty and content are adjusted in real time according to personal performance and ability states of students through dynamic task allocation and personalized learning path optimization; the self-adaptive teaching mode enables each student to learn at proper rhythm and difficulty, a project management tool, a version control system and a real-time feedback mechanism are integrated, the students can obtain real-time guidance and optimization suggestions in actual operation, real-time operation suggestions are provided during code submission and version merging, and the teaching efficiency is improved. And the version snapshot is automatically generated to be associated with the task progress, so that students continuously optimize own operations in practice, automatically generate learning reports, evaluate code quality, task completion efficiency and version control rationality of the students, provide comprehensive learning feedback for the students, and provide data support for teachers to improve teaching strategies.

A stepwise incremental pre-training for integrating discriminative, restorative, and adversarial learning in an AI model. Exemplary systems include means for receiving a training dataset for training a unified AI model. The unified AI model applies each of discriminative, restorative, and adversarial learning operations through three transferable components: a discriminative encoder, a restorative decoder, and an adversarial encoder. Stepwise incremental pre-training operations train the unified AI model, including pre-training the discriminative encoder via discriminative learning and attaching the trained discriminative encoder with the restorative decoder to form a skip-connected encoder-decoder, pre-training the skip-connected encoder-decoder via joint discriminative and restorative learning, and associating the pre-trained skip-connected encoder-decoder with the adversarial encoder. Training of the AI model is finalized by performing full discriminative, restorative, and adversarial learning on the training dataset using the unified AI model.

The present invention relates to a semi-supervised fatigue test state monitoring method based on adaptive confidence active learning, which belongs to the field of equipment state monitoring technology and solves the problems of high labeling cost and poor adaptability to working conditions in traditional monitoring methods. First, a data set consisting of fatigue test state monitoring data is constructed and divided into an initial training set, an active learning set and a validation set. Then, an LSTM is used to construct a state recognition model, and the initial training set is used for pre-training. After pre-training, an active learning model is constructed by adding a Softmax function, and the confidence of the active learning set samples is calculated. Samples below the threshold are screened out, and a retraining set is formed after manual labeling. The retraining set is used to train the state recognition model and adjust hyperparameters, and finally the model is evaluated on the validation set. By integrating active learning with adaptive confidence and a semi-supervised learning mechanism, the present invention significantly reduces the dependence on labeled data, has lower labeling costs, and has strong adaptability to complex working conditions.

This set provides an efficient scale sight-reading learning tool for piano beginners, allowing them to easily learn how to read musical scales on the staff and the correct key placement, even without a piano. [Solution] The scale reading learning set 1, which has a main body with four octaves of musical staff lines, is equipped with rotatably supported sheet winding knobs 4 and 5, and a transparent sheet for writing musical notes 6 that is stretched over the sheet winding knobs 4 and 5 via a transparent sheet for writing musical notes retraction hole 7. By rotating the sheet winding knobs 4 and 5, the musical note marks written on the transparent sheet for writing musical notes 6 move along the musical staff lines. This makes the scale reading learning set 1 compact, limits the range of the musical staff lines to four octaves, and improves the efficiency of learning to read musical notes by allowing the transparent sheet for writing musical notes 6 to move smoothly from side to side and allowing the written musical notes to be read immediately.

The invention discloses an empirical sound velocity error modeling compensation method based on a neural network, which belongs to the technical field of ocean observation data processing, is used for sound velocity error modeling compensation, and comprises the following steps: carrying out data alignment on observation data of CTD and SVP; comparing and analyzing the difference between the sound velocity value calculated by the CTD through an empirical formula and the sound velocity value directly measured by the SVP, and evaluating the correlation through a Pearson's correlation coefficient; constructing a nonlinear regression model by adopting a machine learning algorithm, and continuously optimizing model parameters; establishing a model evaluation index to evaluate the compensated sound velocity precision; machine learning setting parameters are fed back and corrected, and the compensation model is continuously optimized. Compared with the prior art, the method has the advantages that the accuracy of CTD sound velocity calculation is improved, and the established mathematical model has universality and expandability and can be adjusted and optimized according to different sea areas and different measurement conditions; the marine observation cost can be reduced, and the utilization efficiency of the marine observation data can be improved.

The invention relates to the technical field of power station boiler temperature field monitoring, in particular to a power station boiler temperature field online monitoring method and system based on a convolutional neural network and a flame radiation image.The monitoring method comprises the steps that a blackbody furnace is used for calibrating a detector, and the relation between image intensity and radiation intensity is established; a calibrated detector is used for collecting a flame image in the 20%-100% variable load period of the boiler load, and an original flame image is obtained; adding a random shielding object to a flame area in the original flame image to obtain a shielding flame image; and the condition that the detector lens is blocked by slag is simulated. According to the invention, the computer deep learning image restoration technology is introduced into the field of combustion engineering, the non-occlusion flame image and the occlusion flame image are used as the learning set of the convolutional neural network model, and the convolutional neural network model is trained, so that the trained model can process the slag-bonding flame image and output the non-slag-bonding flame image. The problem of lens slagging interference is solved, and the flame temperature field can be accurately measured.

The application discloses a method for expanding random telegraph noise (RTN) based on a storage neural network, and the signal expansion process is realized based on an artificial neural network of a novel storage unit. According to partial RTN measured signals as a learning set, the expansion process of signal prediction reasoning can realize expansion of the signals with an arbitrary time length, and accelerates extraction of the time parameters of the RTN. The method has important significance for development of a physical unclonable function (PUF) technology based on the RTN and information data security.

The invention relates to the technical field of intelligent operation and maintenance of a power system, in particular to an extra-high voltage transformer substation handover test data fusion and abnormity early warning method, which comprises the following steps: collecting and preprocessing multi-source heterogeneous original data to generate a standardized test data set; based on a pre-trained multi-dimensional association rule model, hidden association rules among the test parameters are mined, weighted fusion is executed, and fusion data representing the overall health state of the equipment are generated; comparing the fusion data with a dynamic threshold interval generated by a historical normal sample, calculating a deviation degree and mapping the deviation degree into an abnormal confidence degree; and triggering visual early warning signals of different levels according to the abnormal confidence coefficient, and feeding back a new sample to the learning set for iteratively optimizing the multi-dimensional association rule model and the dynamic threshold interval. According to the method, deep fusion analysis of test data, dynamic quantitative evaluation of the health state and continuous optimization of the model can be realized, and the accuracy of anomaly recognition and the adaptability of the system are improved.

The invention discloses an end-to-end autonomous motion decision method based on deep reinforcement learning. The method relates to the fields of multi-robot motion, machine learning, integrated decision and the like. The method comprises the following steps: firstly, realizing accurate identification of a glass product by adopting improved Mamba-YOLOv10; and then, aiming at the data dependence problem of end-to-end autonomous motion, proposing that an empirical playback mechanism HER and a designed reward function are applied to a deep Q network to realize an HER-DQN integrated decision framework. Compared with other methods, the end-to-end autonomous motion decision-making method not only can accurately identify glass products under complex illumination conditions and environmental interference, but also can improve the stability and robustness of the biochemical laboratory transfer robot during autonomous motion decision-making, and the designed end-to-end autonomous motion decision-making method has real-time performance and high efficiency, and is suitable for popularization and application. The transfer robot in the biochemical laboratory can safely and autonomously complete execution tasks, and the transfer robot can be applied to the fields of agriculture, industry, service industry and the like and has beneficial effects.

This application provides a federated learning contribution evaluation method and device. Multiple edge computing devices are grouped to obtain multiple sub-federated learning sets. For a target sub-federated learning set, model update information corresponding to each edge computing device in the target sub-federated learning set is aggregated. Based on the performance metrics of the updated global model on a public test set, the collaborative contribution value of the target sub-federated learning set is determined. Using a first linear programming solver, the maximum loss value corresponding to all sub-federated learning sets is obtained based on the collaborative contribution value of the multiple sub-federated learning sets, and the maximum loss value is optimized to obtain the minimized maximum loss value. Using a second linear programming solver, based on the minimized maximum loss value of all sub-federated learning sets and the reference contribution value corresponding to the multiple edge computing devices, the target contribution vector corresponding to each of the multiple edge computing devices is determined, accurately quantifying the contribution of each edge computing device in the training process.

A method for setting up an apparatus (1) for biological processes (3), in which process parameters are specified for a plurality of biological processes (3) with computer assistance, that for each biological process (3) a process state is automatically captured, that the particular process state is evaluated using a specified objective with computer assistance, and that from the evaluations the apparatus (1) is set up, with computer assistance, through specification of learned set-up parameters. In addition, an apparatus (1) for biological processes (3) is provided with which the proposed method can be carried out in a particularly advantageous manner.

A method for an object of interest in a degraded 2D digital image of the object is provided. The method includes the following steps: detecting, beforehand, the object of interest in a 2D digital image and assigning it a label; reconstructing a 3D volume of the object thus labeled from a plurality of available 2D digital images of the object of interest; storing, in a database, a record relating to the object thus reconstructed in 3D form and labeled; for each record thus stored, generating a new plurality of 2D digital images according to a plurality of viewing modes from the thus reconstructed 3D volume of each object; training a neural network on a learning set composed of an expanded set of 2D digital images thus generated and corresponding with the label of the object of interest to be recognized; from a degraded 2D digital image of the object of interest to be recognized; using the neural network thus trained to deliver as output the label of the object and a confidence index linked to the recognition of the object of interest.

The invention discloses a curative effect prediction method based on multi-organ metastasis genome data, and belongs to the technical field of medical models, and the method specifically comprises the following steps: collecting clinical pathological characteristics, multi-organ metastasis genome data and a treatment scheme of a breast cancer patient, and recording a metastasis part and a load state; dimensionality reduction is conducted on high-dimensional genome data through a regularization algorithm, feature importance is evaluated in combination with a nonlinear model, and clinical, treatment and genome features related to treatment response are screened out; inputting the screened features into a machine learning and deep learning framework, randomly dividing a training set and a test set in a layered manner, optimizing hyper-parameters through cross validation, and constructing a classic machine learning set model and a deep learning model based on an attention mechanism; disturbing test queue treatment scheme data, evaluating the consistency of model recommendation and an actual scheme, and verifying the prediction capability and clinical practicability of the model; according to the method, multi-dimensional data are integrated, and the curative effect prediction accuracy of the metastatic breast cancer is improved.

The present invention discloses a shard-based content recommendation method, device, and medium, belonging to the field of recommendation methods, including the steps: First, after preprocessing the sample data, a shallow learning integration algorithm based on a traditional single model is built. The output of the decision tree leaf node and the original features are input into the shallow learning network. In the shallow learning network, continuous value data is input into the decision tree model for feature encoding, and cross features are output. Then, the preprocessed categorical features are spliced onto the cross features as the input of ridge regression, and low-order cross features are output. Second, the continuous features in the material data are normalized through deep learning, and the discrete features are transformed into embedding vectors. After splicing the two, they are input into the hidden layer, and high-order cross features are output. Finally, the low-order cross features reflecting the second-order non-linear relationship are spliced with the high-order cross features output by the deep learning network, and the content recommendation result is output. The present invention improves the accuracy of content recommendation.

This invention discloses a method for dynamic compensation of polarization signals in low-altitude complex terrain, an unmanned aerial vehicle (UAV) communication device, and a system, belonging to the field of wireless communication. The method includes: utilizing an edge computing platform to collect and analyze channel polarization state data in real time; employing an adaptive dynamic compensation algorithm; setting a set of candidate modulation schemes based on the channel's signal-to-noise ratio; and constructing a constraint function based on the set of candidate modulation schemes to achieve high-fidelity transmission of polarization signals. The system also integrates interference prediction, federated learning cluster collaboration, and secure communication modules, thereby achieving high security and robust communication while ensuring low latency and low power consumption. This invention effectively solves the problems of polarization signal mismatch, weak anti-interference capability, and poor environmental adaptability in existing technologies under complex terrain, improving the reliability and security of UAV communication systems, and has significant application value and market prospects.

A method for processing distributed data of mobile devices in federated learning based on the Shapley value. Multiple mobile devices are used to construct a federated learning cluster. In each round of federated learning, the central node applies the Monte-Carlo sampling method to estimate the current federated Shapley value of each federated learning mobile device, and takes the projection of the value in the direction of the change of the global model parameters relative to the initial parameters as its importance and contribution to the model. Selecting federated learning mobile devices to participate in the model training of this round based on the federated Shapley value can effectively accelerate the model convergence speed and improve the final accuracy of the model. The present invention can measure the influence of the data sets of each mobile terminal on the model training process, so as to select devices with high contribution degrees to participate in the training in each round, reduce the data communication overhead, accelerate the convergence speed, and improve the model performance.