A recommended model training method and related device

Abstract

An embodiment of the present invention provides a recommendation model training method and a related device, the method comprising: selecting a positive sample in a sample set and adding it to the training set; wherein, the sample set includes a positive sample and a negative sample, and each sample is composed of Composed of n sample features, where n≥1, the sample features of each sample include features used to characterize whether the sample is a positive sample or a negative sample; a preset is used for each multiple negative samples in the sample set The algorithm calculates the sampling probability; selects a negative sample from the sample set according to the sampling probability and adds it to the training set; trains the samples in the training set to obtain a recommendation model. Adopting the embodiment of the present invention in the field of artificial intelligence (AI) for model training can improve the performance of the model.

Description

technical field [0001] The present invention relates to the field of big data technology, in particular to a recommendation model training method and related devices. Background technique [0002] In many machine learning applications (such as recommendation systems) under the background of big data, sampling technology can improve the quality of data training, reduce computing resources, improve training efficiency, and solve the problem of data imbalance, so sampling technology is very important. For example, in the recommendation system of Huawei AppGallery, tens of billions of data are generated every day under user operations. If sampling technology is not used in the training process of these tens of billions of data, the following problems will occur: (1) Due to the large amount of data and too much homogeneous data, the trained model is quickly overfitted; (2) Due to the large amount of data, a large amount of hardware resources are consumed, and the model training i...

AI Technical Summary

Problems solved by technology

For example, in the recommendation system of Huawei AppGallery, tens of billions of data are generated every day under user operations. If sampling technology is not used in the training process of these tens of billions of data, the following problems will arise: (1) Due to the large amount of data and too much homogeneous data, the trained model is quickly overfitted; (2) Due to the large amount of data, a large amount of hardware resources are consumed, and the model training is too slow, which leads to model update If it is not timely, the prediction accuracy of the model will be reduced; (3) If the negative samples are far more than the positive samples, it will lead to serious data imbalance, which will affect the global classification performance of the model

Images