Mini-GD method based on GBP dimension

Abstract

The invention discloses a mini-GD method based on a GBP dimension. According to the method, the problem that a recommendation method in the prior art cannot be combined with gene dimensions for learning, a large amount of memory is occupied during calculation learning; an oscillation value is large in optimization, and overall gradient optimization cannot be carried out during data batch operationare solved, gradient optimization is carried out on data based on a minimum gradient descent algorithm of the GBP dimension, according to the fact that each iteration must advance towards the optimaldirection, meanwhile, no too much memory is occupied during sample training, batch training and learning are carried out, and the optimal solution is guaranteed.

Description

technical field [0001] The present invention belongs to the technical field of data mining, and in particular relates to a mini-GD method based on the GBP dimension. The above-mentioned recommendation method can solve the problem that the recommendation method in the prior art has no way to combine the gene dimension for learning, and takes up a large amount of memory during calculation and learning. When optimizing, the oscillating value is too large, and the overall gradient optimization cannot be performed when the data is calculated in batches. Background technique [0002] With the rapid development of e-commerce websites, recommendation systems have been widely studied and applied. The recommendation system obtains user preferences by extracting and analyzing user information, behaviors, ratings, etc., to help e-commerce companies find specific users for whom it is possible to recommend. Purchased products to increase sales of merchandise. [0003] The recommendation ...

AI Technical Summary

Problems solved by technology

[0007] The technical problem to be solved by the present invention is to provide a recommendation method based on the minimum gradient descent algorithm of the GBP dimension in view of the deficiencies in the above-mentioned prior art. The above-mentioned recommendation method can solve the problem that there is no way to combine the gene dimension For learning, a large amount of memory is occupied during calculation and learning. When optimizing, the oscillating value is too large. When the data is calculated in batches, the overall gradient optimization cannot be performed. This algorithm is based on the minimum gradient descent algorithm. Gene), behavior (Behavior), and phenotype (Phenotypic) are called the user's GBP data, and these data are tagged to form a user-based GBP tag; the user's GBP information is tagged Finally, the content (commodity, article, video, picture, etc.) information is tagged, and then the user's GBP tag and the content tag are matched at the algorithm level; the above algorithm is referred to as a GBP-based mini- GD algorithm, gradient optimization of data

Images