Article similarity calculating method based on restricted Boltzmann machine
A technology of limited Boltzmann machine and limited Boltzmann machine, applied in the field of recommendation system, can solve the problems of lack of item relevance and missing items, etc.
Active Publication Date: 2018-09-14
南方电网互联网服务有限公司
View PDF5 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
The disadvantage of this calculation method is that it only considers the relationship between items from a first-order point of view, and lacks an explanation of the relevance of items from multiple perspectives, and for the sake of calculation speed, this type of similarity calculation method forces The similarity matrix of items is sparse, so a part of the relationship between items is lost
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment 1
[0040] This embodiment provides a similarity calculation method based on a restricted Boltzmann machine to perform multiple angle calculations on items. A Restricted Boltzmann Machine is a parametric generative model with a two-layer structure, such as figure 1 As shown, the lower layer is called the visible layer, and the upper layer is called the hidden layer, with two-way full connection between layers and no connection in the layer. Restricted Boltzmann machines can also be called stochastic neural network models or undirected probabilistic graphical models.
[0041] In a restricted Boltzmann machine, each unit in the visible layer represents an observed variable value, and the units in the hidden layer are used to model the interdependence between units in the visible layer. The entire restricted Boltzmann machine represents a joint probability distribution of visible layer variables and hidden layer variables.
[0042] In the method for calculating item similarity in thi...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention provides an article similarity calculating method based on a restricted Boltzmann machine. The method comprises steps that S1, an article score record of a user in a system is collected,and a user-article score matrix is established; S2, hyper-parameters of a model are set according to calculation tasks, and the super-parameters comprise the aspect number K of article similarity calculation and the number of characteristic values of the corresponding aspects; S3, a restricted Boltzmann machine model is constructed for each aspect, and user score record vectors of articles are utilized as the input of the model to train the restricted Boltzmann machine to acquire a characteristic value activation probability vector value of each article at each aspect and the weight lambdak occupied by each aspect in similarity measurement; and S4, the similarity between a main article and other articles is calculated according to a characteristic value activation probability vector valueand the weight lambdak of the main article at each aspect . The method is advantaged in that the similarity between the articles is calculated through the article similarity calculation method from the multiple aspects, and the calculation result is more accurate and is more consistent with demands of the user.
Description
technical field [0001] The technical field related to the present invention is a recommendation system, and specifically relates to a calculation method for item similarity in a recommendation algorithm. Background technique [0002] In reality, each user has his own viewing angle or focus of understanding for items, so different items will be related to each other from different angles. In other words, the similarity between items exists in Different angles of inspection or focus of understanding. [0003] For example, if you look at the movie "Avatar" from a gender perspective, the movie "Avatar" may be similar to "The Expendables" because it has a lot of fighting scenes, so for men, the two movies may be more similar in this perspective a little. But if you start from the theme of the whole film, the movie "Avatar" and "The Expendables" will not have much similarity, and "Avatar" will be very similar to "The Matrix" in the theme of the film. Therefore, judging from the...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More IPC IPC(8): G06Q30/06G06N99/00
CPCG06Q30/0631G06N3/08G06N3/047G06N3/048G06N3/044
Inventor王国军龙益农刘湘勇邢萧飞
Owner南方电网互联网服务有限公司