A method to improve the stability of collaborative filtering model
A collaborative filtering model and stability technology, applied in instrumentation, computing, electrical digital data processing, etc., can solve the problem of the robustness of recommendation algorithms and other problems, and achieve the effect of improving the robustness of the algorithm
Active Publication Date: 2022-07-26
ZHEJIANG UNIV OF TECH
View PDF5 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0003] But basically all current research is focused on how to improve the accuracy of recommendation, and there are only a handful of research on the robustness of recommendation algorithms.
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 Construction
[0039] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, and the protection scope of the present invention can be more clearly defined.
[0040] refer to figure 1 and figure 2 , a method for improving the stability of a collaborative filtering model, comprising the following steps:
[0041] Step 1: Train the dynamic aesthetic collaborative filtering recommendation system model DCFA to obtain the convergence parameters;
[0042] The prediction model of DCFA is:
[0043]
[0044] where U, V, T are the embedding matrices of users, products and time, respectively, M is the user preference matrix, W is the product latent feature matrix, N is the preference matrix at time r, F is the feature matrix, u, i, r Represent users, items, and time, respectively;
[0045] Step 2: Constru...
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
A method to improve the stability of collaborative filtering models. Since the visual appearance of products has an increasingly important influence on consumers' choices, product visual features have received more and more attention in recommendation. The aesthetic features extracted by the new network structure BDN (Brain-inspired Deep Network) can have better recommendation effect for the recommendation system than the CNN features extracted from the traditional convolutional neural network to represent the image. However, no scholars have studied the influence of aesthetic features on the robustness of the recommendation algorithm. Therefore, in the present invention, we mainly focus on whether the time-varying aesthetic factors will affect the robustness of the model. By adding adversarial perturbation to the embedded matrix of the dynamic aesthetic factors of the model, we found that it will reduce the robustness of the model. Under this premise, we use adversarial training to improve the model and enhance the robustness of the original model. and generalization ability.
Description
technical field [0001] The invention relates to the field of improving the stability of a recommendation system, adopts the currently relatively mature platform Tensorflow to improve the robustness of the model, and specifically relates to a method for improving the stability of a collaborative filtering model. Background technique [0002] Existing research on recommendation technology, some consider different influencing factors to build new algorithms according to application scenarios, and some use new fusion methods to optimize algorithms. The unified space is realized, and the result list of different features is used in the later stage, and the candidate results are used for fusion realization, which effectively integrates the features in different forms to improve the performance of the algorithm. Others optimize the algorithm indirectly by proposing novel feature extraction techniques to optimize the quality of the resulting features. [0003] But basically all 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 Patent Type & Authority Patents(China)
IPC IPC(8): G06F16/9535
CPCG06F16/9535
Inventor 吴哲夫李泽农
Owner ZHEJIANG UNIV OF TECH