Unsupervised learning unified feature extractor construction method

Abstract

The invention provides a constructing method for an unsupervised learning unified characteristic extractor. The constructing method is characterized in that actual news text data is acquired from a server-side to generate a news characteristic training data set; data of the news characteristic training data set is processed and quantized to obtain a news characteristic training vector set; according to user accessing data, news data sets are classified to generate a user characteristic training data set; a stacked asymmetrical noise reduction contraction autoencoder with a plurality of implictstrata is constructed, and a certain objective function is used to train a deep autoencoder; after the training of the deep autoencoder is completed, a decoder portion is deleted, a binaryzation generating layer is added to construct the unsupervised learning unified characteristic extractor. According to the unsupervised learning unified characteristic extractor, the unification of news characteristics and user characteristics and the unification of content recommendation and collaborative filtering recommendation can be achieved, and the real-time recommendation efficiency is improved.

Description

technical field [0001] The invention belongs to the field of artificial intelligence, and in particular relates to a construction method of an unsupervised learning unified feature extractor. Background technique [0002] The current recommendation system or recommendation engine is generally divided into content-based recommendation, collaborative filtering recommendation, and hybrid recommendation. The current popular collaborative filtering method is mainly based on commonality, that is, the similarity between users and items is calculated through a user's rating of a product or media content (which can be collectively referred to as "item"), and then according to the user's interest Similar to other users' ratings to infer their ratings for new items, or to predict ratings for new items based on the similarity with items they were interested in, so it is also called rating prediction, but its disadvantages are insufficient personalization, Difficulty predicting in the a...

AI Technical Summary

Problems solved by technology

[0006] In view of the lack of personalization in the current fusion media news recommendation and other applications, the difficulty of extracting user features, the difficulty of unifying different methods to form an effective hybrid recommendation method, the privacy violation in user feature extraction, and the need to improve the efficiency of real-time recommendation, according to the current A new type of artificial intelligence technology, this application discloses a construction method of Unsupervised Learning Unified Feature Extractor (ULUFE), which is used to extract "Unified Representation Based on Content (URBC) )

Images