5 results about "Kernel model" patented technology

The invention discloses a vertical domain question and answer benchmark generation method and device, equipment and a medium, and relates to the technical field of vertical domain question and answer, and the method comprises the steps: obtaining a target theme, and obtaining a target question through a preset vertical domain kernel model according to the target theme; the preset vertical domain kernel model is a pre-trained large language model based on a retriever, and the preset vertical domain kernel model comprises an external knowledge base corresponding to a target vertical domain where the target topic is located; according to a preset context example and a preset first standard cue word, obtaining a general evaluation standard corresponding to the target topic through a preset vertical domain kernel model; according to the target question, the general evaluation standard and a preset second standard cue word, obtaining a refined evaluation standard through a preset vertical domain kernel model; and obtaining a to-be-tested answer according to the target question and the to-be-tested language model, and evaluating the to-be-tested answer according to the refined evaluation criterion to obtain an evaluation result. According to the invention, an effective evaluation method can be provided to evaluate the accuracy and specialty of the to-be-tested model in vertical domain questions and answers.

The application relates to a key indicator trend early warning method and system, a terminal and a storage medium, and belongs to the technical field of data processing. The method comprises the following steps: acquiring original data, a first term coefficient of a convolution kernel and the number of convolution kernels; constructing a convolution kernel model according to the first term coefficient of the convolution kernel and the number of the convolution kernels; scanning the original data by using the convolution kernel model to obtain a scanning interface, and marking local original data located in the scanning interface as a target vector; performing bucketing on the target vector to obtain bucketing characteristic values; calculating the bucketing characteristic values adjacent to each other by using a first-order difference function to obtain a difference vector; calculating a trend probability of the scanning interface according to the difference vector, qualitatively determining the trend of the scanning interface; and calculating a trend growth rate of the scanning interface according to the bucketing characteristic values, quantitatively determining the trend of the scanning interface. The application can achieve the purpose of intelligently operating and maintaining industrial equipment.

A data distillation system (1) generates, from a dataset used for training a kernel model, a synthetic dataset having a smaller number of pieces of data than the number of pieces of data constituting the dataset. The data distillation system comprises an information acquisition unit (20) and an information processing unit (30). The information acquisition unit acquires a dataset, a first model parameter obtained by training a kernel model using the dataset, and a synthetic data count indicating the number of the synthetic datasets generated. It is assumed that a model parameter obtained by training the kernel model using the synthetic dataset of the synthetic data count is a second model parameter. In this case, the information processing unit defines an upper bound value of a variation range of the second model parameter with respect to the first model parameter by using a predetermined upper bound evaluation method, and generates the synthetic dataset through processing for minimizing the upper bound value.

In a decision-support system for gastrointestinal (GI) endoscopy, convolutional neural networks (CNNs) are set up to perform decision-support tasks according to endoscopic images. Each learnable kernel used in the CNNs is advantageously modeled as a linear combination of a set of fixed kernels for simplifying kernel learning, giving a lightweight kernel model to advantageously reduce required computation resources. Further computation-resource reduction can be made by CNN model compression via knowledge distillation and by using multi-task CNNs. It enables the decision-support system to be realized as an edge computing system near a site of performing endoscopic examinations. The system can be automatically configured for esophagogastroduodenoscopy (EGD) or colonoscopy. In the system, lesion-detection results and quality-control results can be seamlessly integrated to provide value-added results, which are more valuable to the endoscopist than separately considering the lesion-detection results and quality-control results.

The invention discloses an E-PINN-based sintering ignition temperature prediction method, and belongs to the technical field of sintering control. Comprising the following steps: 1, measuring operation parameters of the ignition furnace, forming input features and output features, processing data of the input features and the output features, and dividing the data into a training set and a test set; 2, carrying out pre-training on the training set by adopting ridge regression with L2 regularization to obtain a regression coefficient and a bias coefficient of a linear physical kernel model, and establishing the linear physical kernel model according to the regression coefficient and the bias coefficient; and 3, introducing a residual network to compensate the linear physical kernel model to obtain an ignition temperature prediction model. According to the method, the data fitting error and the physical consistency constraint can be unified into a training target, and the unification of the prediction precision, the stability and the physical consistency is realized under a complex working condition.