162 results about "Process training" patented technology

The invention provides a method and a system for detecting DNS (domain name system) traffic abnormality. The method includes: extracting corresponding characteristic values for DNS traffic data to be processed, giving different weights to each characteristic, and detecting an abnormality cluster marked in a training set by the aid of the W-Kmeans algorithm and the additional Euclidean distance threshold Dthreshold so that new unknown characteristic abnormality can be discovered. The method and the system have the advantages that the algorithm is high in convergence speed and small in calculation, new samples to be detected only need to be compared with a processed training clustering center, calculation of a great quantity of original training data is not needed, the method and the system are low in deployment cost, strong in generalization ability and capable of discovering DNS traffic abnormality rapidly and effectively, and the system is particularly suitable for being deployed on a large DNS server.

The invention discloses a dynamic confirming method for performance index value normal fluctuating range, comprising following steps: obtaining a historical value of the network performance index, and dynamically obtaining the latest historical sample data; processing pre-treatment to obtain a normal sample data; processing phase space re-construction to obtain a training sample data; processing training modeling to take residual error white noise as condition of selecting the optimal supporting vector machine model; using the optimal supporting vector machine model to predict the data on the to-be predicted time point, and calculating the confidence interval of the predicted value to obtain the performance index value normal fluctuating range; detecting whether the optimal supporting vector machine model is suitable for the next prediction of the to-be predicted time point; if the model is not suitable, processing re-training. The invention also discloses a device for the dynamic confirmation performance index value normal fluctuating range. The invention significantly enhances the precision of the performance dynamic early warning, and reduces wrong report and missing report of the performance warning.

This invention relates a control strategy for a hybrid electric vehicle having an electric motor, a battery and an internal combustion engine. The control strategy improves fuel economy and reduces emissions while providing sufficient acceleration over a varying set of driving conditions through an adaptive control unit with an artificial neural network. The artificial neural network is trained on a pre-processed training set based on the highest fuel economies of multiple control strategies and multiple driving profiles. Training the artificial neural network includes a training algorithm and a learning algorithm. The invention also includes a method of operating a hybrid electric vehicle with an adaptive control strategy using an artificial neural network.

The invention provides a depth neural network based vector quantization system and method, comprising a normalization preprocessing module for normalizing original data through normalized data and outputting preprocessed data after normalization; a vector quantization and coding module for receiving the preprocessed data and the codebook and carrying out vector quantization coding to the preprocessed data through the codebook and outputting the coded data; a neural network inverse quantization module for performing the decoding of the inverse quantization to the coded data through a depth neural network and outputting the decoded data; a processing module after inverse normalization for performing an inverse normalization process to the decoded data through the normalized data and outputting the restored original data after the inverse normalization; and a neural network training module for carrying out trainings to the neural network through the pre-processed training data and coded training data after normalization processing and outputting the neural network to the neural network inverse quantization module. The system and the method of the invention can effectively solve the problem that the quantization error is large in high dimension signal vector quantization.

The present invention provides a random voiceprint certification system comprises a training system, a random cipher generator, and a testing system, which is employed to process training or testing operation for the input raw voice data. In training voice, the training system obtains an appointment voiceprint feature model parameter groups from the input raw voice data. From the appointment voiceprint feature model parameter groups several voiceprint characteristic units are obtained and at least one reference voiceprint password, which is for the testing system to carry out the voice testing operation is built. In processing testing voice, the random cipher generator generates randomly at least one reference voiceprint password from the voiceprint characteristic units of the appointment voiceprint feature model parameter groups to build the random voiceprint cipher lock. The present invention generates randomly one or several reference voiceprint passwords. The random voiceprint certification system is built completely to form the random voiceprint cipher lock. Therefore, the effect of not easy for illegal invasion can be achieved.

The invention discloses an ore scale measurement method based on deep learning and an application system. The method comprises the following steps: obtaining an ore block image; image preprocessing: processing the implemented ore block image into a marked image, and dividing the marked image subjected to processing into training sample boxes and test samples; removing the abnormal marked image data; training a preset RetinaNet target recognition network by using the processed training sample; inputting the test sample to a target identification network to obtain a target identification result,and calculating the size of the ore; the invention discloses an ore scale measurement method based on deep learning and an application system. The RetinaNet target recognition network is trained by adopting a labeled ore image sample; the trained network model is obtained to be used for classifying and positioning the ore blocks, the real sizes of the ore blocks are calculated, complex features do not need to be extracted manually, the detection efficiency is high, and the problem that efficiency is low in traditional ore scale measurement is solved.

The invention discloses a power grid load prediction method based on a depth LSTM neural network, and the prediction precision, step length and the real-time performance can be improved. The method comprise the steps of generating a training sample according to input characteristic data and the load data, wherein The input characteristic data comprises meteorological information of experimental time and time type information of whether it is working days, processing the training samples, training the processed training samples through the LSTM neural network to obtain an LSTM prediction model,inputting the meteorological information of experimental time and time type information of whether it is working days, into the LSTM prediction model to predict the power grid load in the to-be-predicted time so as to obtain a power grid load prediction result, analyzing the power grid load prediction result to determine whether the power grid load prediction result meets the accuracy requirement; if the accuracy requirement is not met, obtaining new training samples, and carrying out supplementary training on the LSTM prediction model through the new training samples so as to update the LSTMprediction model.

The invention discloses a network index prediction method and device based on ARIMA model and a storage medium. The method comprises the steps of: collecting index data of a to-be-predicted index variable in a certain time period as a training data set; taking the pre-processed training data set as input to construct an ARIMA model of network index prediction; and commonly inputting the future time length of the set to-be-predicted index variable and the selected stationary series having passing d-order difference into the ARIMA model, and calculating and obtaining a target prediction value. The change rules of users for wireless network demands are subjected to statistics to predict the change sequence value of an index in a certain time in the future so as to provide more effective datareference for optimizing wireless network resource allocation and performance optimization.

The invention relates to an attention mechanism-based image aesthetics quality evaluation method, and belongs to the technical field of computer vision. The method comprises the steps of firstly processing training data, then designing a network structure model, adopting a lightweight deep network as a backbone network, and integrating the backbone network into an attention mechanism module; designing a loss function for training the network based on a data equalization thought; finally, training a network structure model by using the processed training data to obtain a network model capable of automatically evaluating the aesthetic quality of an image; and performing aesthetic scoring on the input picture based on the model, and applying the model to shooting to assist a user to shoot a more beautiful picture in real time. Compared with the prior art, the network structure model adopted by the invention can more effectively extract the characteristics of images, the adopted loss function greatly enhances the data learning ability of the model. Compared with other methods, the accuracy is improved, and the parameter quantity of the model is reduced.

A method of processing training data is provided. The method includes accessing training data stored in at least one of a plurality of training databases, the training data being related to at least one of a plurality of employees. The method also includes providing an on-line training course for use by at least one of the plurality of employees, the on-line training course being distinct from the plurality of training databases. Additionally, the method includes formatting the training data included in the at least one of the plurality of databases such that the training data can be processed by a master database. The method also includes importing the formatted training data into the master database. Further, the method includes updating the master database to include training information related to the on-line training course upon the successful completion of the on-line training course by one of the plurality of employees is provided.

The invention discloses a gesture recognition method based on an improved residual neural network. The method includes the following steps: S1, acquisition of a training sample set; S2, preprocessingon the training sample set, wherein positions of gestures in images are found through algorithms, and cropped images are used as original training data; S3, enhancement of training samples, wherein translation transformation, rotation transformation, mirror-image transformation, scaling transformation and the like are carried out on the collected training samples to enlarge the training sample set; S4, acquisition of a gesture model, wherein a processed training sample set is input into the pretrained residual network to carry out training on network parameters to obtain the gesture recognition model; S5, a step of carrying out processing, which is the same as the step S2, on to-be-recognized gesture images to obtain to-be-recognized gesture data; and S6, a step of inputting the to-be-recognized gesture data into the network, of which training is completed, to obtain a gesture sequence. The method is based on the deep residual network, trains the residual network on the self-collecteddata set, and realizes high-recognition-rate gesture recognition of a third view angle.

The present invention relates to a fault diagnosis method based on a metric learning and time sequence during an industrial process for solving the problems that a conventional fault diagnosis method is high in system cost, is difficult for on-line diagnosis, is difficult to distinguish to the fault types, etc. The method is realized by a step 1 of dividing the system faults into n types; a step 2 of preparing a training sample; a step 3 of pre-processing the training sample; a step 4 of carrying out the metric learning on the pre-processed training sample; a step 5 of calculating the distances between a real-time monitoring sample and n subclasses; a step 6 of according to the distances between the real-time monitoring sample and the n subclasses, adopting a KNN classification method to diagnose the faults, namely determining whether a system generates the faults and determining the types of the faults. The fault diagnosis method based on the metric learning and time sequence during the industrial process of the present invention is applied to the fault diagnosis field.

The invention discloses a moving target detection method based on a deep optical flow and a morphological method, which includes the following steps: (1) collecting video data, marking sample videos,randomly dividing the sample videos into a training set and a testing set, carrying out mean calculation on the processed training set and the processed testing set to form a training set mean file and a testing set mean file, and completing the preprocessing of the training set and the testing set; (2) constructing a fully convolutional neural network architecture composed of a coding part and adecoding part, and carrying out training by using the training set and the testing set through an adaptive learning rate adjustment algorithm to get trained model parameters; (3) inputting image dataneeding detection to a trained fully convolutional neural network to get a corresponding deep optical flow graph; (4) processing the deep optical flow graph through an Otsu threshold adaptive segmentation method; and (5) morphologically processing the data after threshold segmentation, removing outliers and slots, and finally obtaining a detected moving target area.

The invention relates to a safety measure demonstration simulation display method and apparatus for current loop operation of a substation. The method comprises the steps of determining a display caseaccording to a safety measure of normalized current loop operation of the substation; according to the display case, collecting field materials, and according to the field materials, building a substation three-dimensional model required for the whole process of the normalized current loop operation by using specific three-dimensional modeling software; importing the substation three-dimensionalmodel to a virtual reality simulation display platform generated based on a Unity3D virtual reality engine, and based on the substation three-dimensional model, realizing a scene loading and scene resource management function, a scene stereo display and real-time rendering function, a function of man-machine interaction with a virtual reality hardware interaction device, and a safety measure demonstration process simulation display function; and performing safety measure simulation display for the display case on the virtual reality simulation display platform by utilizing a virtual reality simulation display device. Based on a VR technology, various operation process trainings are efficiently finished.

The embodiment of the invention discloses a machine learning model generating method and device, computer equipment, and a storage medium. The method comprises the steps: obtaining training data; calling a pre-stored data processing module and a pre-stored algorithm module; carrying out the data preprocessing of the training data through the pre-stored data processing module, so as to obtain the processed training data; carrying out the training and verification of the pre-stored algorithm module according to the processed training data, so as to obtain a machine learning model and verification indexes corresponding to the machine learning model; and displaying the machine learning model and the verification indexes corresponding to the machine learning model according to a preset displayrule. According to the invention, the machine learning model can be obtained each time without programming, thereby greatly reducing the workload of an engineer, and improving the obtaining efficiencyof the machine learning model.

The invention provides a model generation method, device, a computer device and a storage medium. The method comprises: receiving original training data input by a user, specifying a variable type ofthe original training data, and specifying a training model algorithm; Invoking a pre-stored data preprocessing method corresponding to the variable type; using the data preprocessing method to preprocess the original training data to obtain the processed training data; Invoking the training model algorithm; machine learning is performed on the training data using the processed training model algorithm to generate a machine learning model. A method for pre-store data corresponding to different parameter types Given a variety of optional initial training models and algorithms, we only need to call appropriate data preprocessing methods to process the data and call appropriate initial training models or algorithms to generate machine learning models, which greatly reduces the workload of engineers and improves the efficiency of acquiring machine learning models.

The invention belongs to the technical field of image processing, and particularly discloses an SAR image target recognition method based on non-negative matrix factorization of sparse constraint. According to the method, more effective features are extracted by improving a non-negative matrix factorization method to improved recognition precision, and the problems that features extracted in the prior art are not typical and recognition precision is not high are solved. According to the method, a training sample image and a test sample image are pre-processed in the same way, logarithmic transformation is carried out, a pre-processed training sample set is factorized through non-negative matrix factorization of sparse constraint to obtain a basis matrix and a coefficient matrix, a test sample set is projected in a sub space of a basis matrix structure, classification is carried out through an SVM after a feature matrix is obtained, and then classification accuracy is obtained ultimately. Compared with the prior art, the extracted features are more effective and recognition precision can be effectively improved.

The invention discloses a substation simulation training system which comprises a normal basic operation module, a protective and automatic device module, and a failure and abnormal accident processing module. The normal basic operation module is used for simulating operable equipment in a substation. The logic relationship of simulating devices is uniform with the logic relationship of actual devices. The simulating devices can be operated in a far mode or a local mode. The protective and automatic module is used for simulating motion behaviors of normal protection of relay protection in an actual system and simulating functions of an automatic device. The failure and abnormal accident processing module is used for simulating various equipment failure types and equipment abnormality types. The substation simulation training system is friendly in human-computer interface, simple, easy to learn, and simple in operation. In the utilization aspect, most operation of the system can be finished through a mouse or shortcut keys. The substation simulation training system brings convenience to replacement of a substation to be simulated, saves investment, enables a user to study on a machine at any time, can conduct accident process training repeatedly, is strong in pertinence and convenient to upgrade, and enables the user to use just after learning.

The invention relates to a machine learning algorithm whole-process training method and system based on a cloud infrastructure. The method comprises the steps that a training data set and a to-be-solved problem data set are uploaded to a cloud data server through a web application program, and a preset machine learning algorithm is selected; the training data set is obtained by a cloud computing server through a cloud data server, and model training is conducted by using the training data set to obtain a training model; the to-be-solved problem data set is obtained by the cloud computing server through the cloud data server, solving results are obtained by using the training model according to the to-be-solved problem data set, and the solving results are returned to the cloud data server;the solving results are returned to the web application program by the cloud data server. According to the method, computing environments and computing resources are deployed at cloud ends to help technical researching and developing workers in efficiently and rapidly constructing a machine learning algorithm model and efficiently training the model, and therefore time cost and software and hardware purchasing cost are reduced.

The invention discloses a weblog anomaly detection method in combination with GRU and SVDD, and the method comprises the steps: carrying out the dimension reduction of a large-scale weblog data set through principal component analysis, and extracting effective attributes; then, using the processed training data set for training a GRU-SVDD classifier model; and finally, inputting an actual log to be detected into the GRU-SVDD comparator, and detecting an exception in the log. The method is superior to classical GRU-MLP, LSTM and other algorithms. According to the method, the abnormal conditionin the test data set can be well detected and positioned.