40results about How to "Uniform data" patented technology

Disclosed is a transformer fault detecting method based on a simplified set unbalanced SVM. The method comprises (1) obtaining a characteristic vector set through a fault characteristic extracting method based on GARCH models; (2) performing determination of boundary samples on minority-class samples to obtain a minority-class boundary sample set S, wherein the minority-class samples are fault samples; randomly selecting N[x]={2, , ISI}, wherein ISI is the cardinal number of S, a[i]=1, and i=1, N[x], and setting N[z] to be 1, utilizing a simplified set solution algorithm to obtain Z[1] and repeating the operation for N[L]-N[M] times, wherein N[L] is the number of majority samples, N[M] is the number of minority samples, and accordingly, N[L]-N[M] is the number of artificial minority samples, and guaranteeing N[z]=ISI for at least one once; (3) combining the artificial minority samples obtained in the step (2) with original minority samples to serve as the training samples of an SVM classifier and lastly to obtain an SVM decision model; (4) inputting newly-obtained transformer characteristic vectors into the decision model for judgment. The transformer fault detecting method based on the simplified set unbalanced SVM is applied to transformer fault detection.

The invention discloses a data link equalization method, a mobile terminal and a computer-readable storage medium. For a mobile terminal including two WiFi antennas and two data channels, when two wireless hotspots are connected at the same time, the two data channels To transmit the data link generated by the upper layer application, the mobile terminal detects the link delay of the two data channels, calculates the average link delay of the two data channels based on the detected link delay and the preset smoothing algorithm, and then According to the average link delay of the two data channels, the data link balance ratio of the two data channels is determined, so that when the new data link of the upper layer application is detected, the upper layer application is newly created based on the determined data link balance ratio. The data link allocation ratio of the two data channels is transmitted to the two data channels, and the data link balance ratio is adjusted through the link delay of the two data channels, which can effectively ensure the data balance of the two data channels and improve the network access efficiency.

A method for dispatching queue in data network includes confirming weight of weighted dispatch for each queue and confirming relevant weighted dispatch interval in a weighted dispatch cycle according to weight of each queue, relevantly dispatching each queue data to be out of queue for being sent out according to confirmed weighted-dispatch interval of each queue. In addition, a relevant queue dispatching device is also disclosed.

The present invention provides a traffic network traffic light control method and system based on edge computing, and a computer-readable storage medium. Information is shared between edge nodes connected to road sections, and a central cloud server shares information of all nodes. Each edge node sends the real-time traffic operation status to the cloud server, and the cloud server optimizes the timing scheme of the entire traffic network according to a specific optimization algorithm to obtain the best setting time for the current traffic lights at each intersection; and then the cloud The server sends the corresponding time adjustment scheme to each edge node. At the same time, each edge node will optimize the regional timing based on the local traffic network formed by the nodes connected to it, and obtain a local timing scheme based on regional traffic optimization; the two schemes will compete, and the optimization goal will be improved the most. will be adopted; then each edge terminal control module will make real-time adjustments to the traffic signal equipment controlled by it.

The invention discloses a small-sample plant disease identification method and system. The method includes: randomly selecting a plurality of original pictures containing diseases as the first sample set; expanding the first sample set through an improved deep convolution-based generative adversarial network to obtain a second sample set; performing a second sample set on the second sample set Verifying and constructing a training set together with all or part of the second sample set that has passed the verification and the original picture that does not contain the disease; training the convolutional neural network based on the training set to obtain a classification model; inputting the image of the disease to be identified into the classification model to obtain the disease Recognition results. In the present invention, aiming at the disease pictures of small samples of plants, the improved deep convolution generation confrontation network is used to expand the sample set, so that the positive and negative ratio of the expanded second sample set is roughly 1:1, so that the data is balanced, and the two The number of patients has reached 10,000, and the convolutional neural network is used to classify the diseases on the expanded second sample set, which has a good classification effect.

The embodiments of this specification provide a method and device for expanding training samples. The method includes: obtaining the initial training sample group to be expanded, including the first number of training samples, the training samples include the original machine sentences and original user sentences in the historical dialogue group, and the category labels corresponding to the dialogue group, and the initial training sample group Each training sample of has a first category label; Obtain a second number of training samples from the initial training sample group; For each dialogue group in the second number of training samples, the first class associated with the original machine sentence in each dialogue group A machine statement is input into the dialog generation model of the pre-trained first category label, and the first user statement corresponding to the first machine statement in each dialogue group is generated; the first machine statement and the corresponding first machine statement in each dialogue group are The user sentence is added to the initial training sample group as an extended training sample to obtain an extended training sample group. Able to achieve data equalization of training samples.