Model training method and device for training data

Abstract

The invention discloses a model training method and device for training data. The method comprises the following steps: obtaining original training data; carrying out aggregation on the original training data to obtain aggregated training data; according to the original training data and the aggregated training data, establishing an index vector, wherein the absolute value of an index vector value is used for indicating the position of the training data in the aggregated training data in the original training data; randomly reading the value of the index vector, and obtaining corresponding training data from the aggregated training data according to the value; and utilizing the obtained training data to carry out model training. On the premise of training data aggregation, the value of the index vector is randomly read, the corresponding training data can be obtained from the aggregated training data, and the randomness of the training data used for the model training is guaranteed so as to improve a model training effect on the basis of memory saving.

Description

technical field [0001] The invention belongs to the technical field of computing, and in particular relates to a model training method and device for training data. Background technique [0002] The click-through rate estimation of online advertisements plays an important role in the process of advertising. At present, the industry mainly uses simple linear models such as logistic regression (LR, Logistic Regression) to model the click-through rate of advertisements. The model solving process is simple and fast, and can be To a certain extent, it prevents overfitting of the data, etc. Although the LR model is simple to solve, in the era of big data, it is still necessary to fully exploit the computing performance of LR. Stochastic Gradient Descent (SGD, Stochastic GradientDescent) algorithm is an optimization algorithm commonly used to train LR models, which can achieve faster convergence in scenarios with massive data. [0003] In the training data generated for advertisi...

AI Technical Summary

Problems solved by technology

[0003] In the training data generated for advertising click rate estimation, there are often a large number of repeated training data, and these large amounts of repeated data waste storage space to a large extent

If these repeated training data are aggregated and only one copy of the training data is kept, although the memory is saved, the same data will be accumulated in one place after aggregation, which will destroy the uniform distribution of the data and fail to satisfy the original randomness of the data. sex

However, the SGD algorithm needs to obtain better model training results on a training data set that guarantees randomness. Training on an aggregated training data set often results in poor model results.

Images