10 results about "Self training" patented technology

The application provides an optimization method, system and equipment for a public and railway combined transport delivery network under external set internal matching, and belongs to the technical field of logistics distribution network optimization. The method is aimed at the problem that the traditional optimization algorithm in the prior art is prone to local optimization and difficult to obtain a globally satisfactory solution, and uses a deep learning assisted genetic algorithm to solve a pre-constructed public and railway combined transport network model. The algorithm performs real-time sensing on the population evolution state through a deep Q network, dynamically and intelligently selects a genetic operation strategy, and combines an experience playback mechanism to perform self-training and parameter updating, and finally outputs an optimal two-stage public and railway combined transport hub site selection and network freight flow distribution scheme. The application uses the above optimization method, system and equipment for a public and railway combined transport delivery network under external set internal matching, effectively improves the global search capability and adaptive level of the algorithm, and can obtain a public and railway combined transport network optimization scheme which is lower in cost, better in layout, and stable and reliable.

This application discloses a method, device, terminal, and medium for controlling power system signal synchronization, relating to the field of power system control technology. The technical solution provided by this application calculates the phase difference and frequency difference between the output signal and the reference signal based on their phase and frequency data, respectively. The phase difference and frequency difference are then input into a signal synchronization optimization model based on an adaptive neural network algorithm, which then outputs a synchronization signal through the operation of the signal synchronization optimization model. The signal synchronization optimization model, constructed using the adaptive neural network algorithm, can continuously update parameters based on the characteristics of the input data, perform self-training, and flexibly adjust to accommodate different frequency and phase drifts, thereby improving the stability and accuracy of power system signal synchronization.

This invention discloses a method and system for intelligent scheduling of unmanned aerial vehicles (UAVs) based on a knowledge-driven meta-learning device. Specifically, the method comprises: deploying and initializing a knowledge-driven meta-learning device, including meta-training and target retraining modules; the meta-training module determines the meta-learning objectives of the UAV intelligent scheduling model and optimizes meta-parameters, performing weight estimation through linear regression and measuring the similarity between different feature representations through a similarity metric; constructing a meta-task environment covering multiple tasks to learn cross-task update rules, using feature domain basis matrices to form orthogonal multi-domain feature representations; introducing a physical guidance term into the meta-learning loss function to align the output with physical laws and optimize the meta-learning objective function; and the target retraining module performs self-training using unlabeled data, updating the parameter set to obtain the final model, and directing and optimizing the collaborative operation of multiple UAVs. This invention can rapidly optimize the performance of the UAV intelligent scheduling model even when data is scarce.

The application discloses a single picture rain removing method based on a memory-oriented Transformer, characterized by an encoder-decoder structure with a self-supervised memory module, which can well process input pictures and better extract required features, wherein the self-supervised memory module is a neural network with memory, which can record various forms of rainfall, wherein each entry in the memory corresponds to a prototype feature of a rain pattern; a self-training mechanism is added to the self-supervised memory module, enhancing the adaptability of the algorithm to natural rain pictures. Compared with the prior art, the application can remove more rain streaks with different appearances, restore clearer background scenes, and better retain the structure and details of the background. The addition of the self-training mechanism makes the algorithm more adaptive to natural rain pictures, and good results can also be achieved on natural rain pictures.

Using processes and methods described herein, a digital twin of a physical space can train itself using sensors and other information available from the building. In some embodiments, a system to be controlled comprises a controller that is connected to sensors. This controller also has a thermodynamic model of the system to be controlled. The thermodynamic model has neurons that represent a thermodynamically coherent section of a building, such as a window. The neurons represent these portions of the controlled space using parameter values and equations that model physical behavior. A machine learning process refines the thermodynamic model by modifying the parameter values of the neurons, using sensor data gathered from the system as behavior to be matched by the thermodynamic model. The thermodynamic model may be warmed up by running the model using state data as input.

The invention discloses a Tri-training-based adaptive rotating machine fault diagnosis method in the passive data field. The method comprises the following steps: S1, carrying out source domain data input and initial three-classifier training; s2, target domain data input and pseudo label generation are carried out; s3, result scoring and model fusion are carried out; s4, performing cooperative training of the three classifiers; s5, carrying out iterative optimization and self-adjustment; and S6, outputting a target domain data fault diagnosis result. By introducing mutual supervision and self-training of the three classifiers, model optimization can be carried out by effectively utilizing unlabeled data. Compared with other adaptive methods in the passive data field, the method utilizes a feedback mechanism of a plurality of models to generate pseudo labels through continuous iteration, so that the utilization efficiency and stability of label-free target domain data are improved, the fault diagnosis precision of the models under the target domain working condition is gradually improved, and the fault diagnosis efficiency is improved. And an efficient and reliable solution is provided for fault diagnosis of the rotating machinery of the power system.

The application discloses a network traffic anomaly detection method based on interactive game, and combines a reinforcement learning algorithm; in a training process, a model receives reward and punishment signals from the outside world through feedback, thereby playing a guiding role in self training, making the interactive behavior between the model guide body and the model reach a dynamic balance state, and thereby controlling the frequency of interaction and feedback. The network traffic data processing method adopts an interactive game mode, the model guide body feeds back and evaluates the analysis and decision of the model according to a model judgment standard, and then the model continuously obtains more accurate training results in the training process according to external information. In order to better make the model guide body and the model cooperate, establish a reasonable interaction strategy of the model and a reasonable feedback strategy of the model, a dynamic Bayesian game model is used to establish a credit evaluation and updating mechanism between the two, and an equilibrium state between the two models is established through game.

Using processes and methods described herein, a digital twin of a physical space can train itself using sensors and other information available from the building. In some embodiments, a system to be controlled comprises a controller that is connected to sensors. This controller also has a thermodynamic model of the system to be controlled within memory associated with the controller. The thermodynamic model has neurons that represent distinct pieces of a controlled space, such as a piece of equipment or a thermodynamically coherent section of a building, such as a window. The neurons represent these distinct pieces of the controlled space using parameter values and equations that model physical behavior of state with reference to the distinct piece of the controlled state. A machine learning process refines the thermodynamic model by modifying the parameter values of the neurons, using sensor data gathered from the system to be controlled as ground truth to be matched by behavior of the thermodynamic model. The thermodynamic model may be warmed up by running the model using state data as input.

To provide a method and an apparatus for synthesizing voices similar to actual voices using an artificial neural network self-learned by self-supervised learning without the need to train the artificial neural network using a large amount of voices and text datasets.SOLUTION: Provided is a self-supervised learning-based voice synthesis method including: training a voice analysis module to output voice features for training voice signals by using the training voice signals representing training voices, and outputting voice features for the training voices; training a voice synthesis module to synthesize voice signals from the voice features for the training voices by using the output voice features, and synthesizing synthesized voice signals, representing synthesized voices, from the output voice features; and calculating reconstruction loss between the training voice signals and the synthesized voice signals, and training the voice analysis module and the voice synthesis module based on the calculated reconstruction loss and the training voices.SELECTED DRAWING: Figure 8

The invention provides an optimization method, system and device for a highway-railway combined transportation delivery network under external collection and internal distribution, and belongs to the technical field of logistics distribution network optimization. According to the method, in order to solve the problems that in the prior art, a traditional optimization algorithm is prone to falling into local optimum, and a global satisfactory solution is difficult to obtain, a deep learning assisted genetic algorithm is adopted to solve a pre-constructed highway-railway combined transportation network model. According to the algorithm, the population evolution state is sensed in real time through the deep Q network, a genetic manipulation strategy is dynamically and intelligently selected, self-training and parameter updating are carried out in combination with an experience playback mechanism, and finally an optimal secondary highway-railway combined transportation hub site selection and network cargo flow allocation scheme is output. According to the optimization method, system and equipment for the highway-railway combined transportation delivery network under the external set and internal distribution, the global search capability and the adaptive level of the algorithm are effectively improved, and a highway-railway combined transportation network optimization scheme which is lower in cost, better in layout, stable and reliable can be obtained.