Back to results
US 12,614,108 B2Granted
Method for recommending information, recommendation server, and storage medium
AssigneeBoe Technology Pte. Ltd.
InventorYongchi Zhang
Life of this patent
Estimating the term…
Abstract
Provided is a method for recommending information. The method includes acquiring a click through rate of recommended information by predicting a feature vector of the recommended information by a recommendation model, wherein model parameters corresponding to features that are the same as features of the feature vector and have frequencies less than or equal to a first threshold are removed from the recommendation model; and issuing the recommended information based on the click through rate.
Claims17
A method for recommending information, executed by a recommendation server, and comprising:
- acquiring a click through rate of recommended information by predicting a feature vector of the recommended information by a recommendation model, wherein model parameters corresponding to features that are the same as features of the feature vector and have frequencies less than or equal to a first threshold are removed from the recommendation model;
- and
- issuing the recommended information based on the click through rate, wherein the recommended information is data content corresponding to a current issue request and is looked up by the recommendation server based on a received issue from a client;
- wherein the recommendation model is generated by a system for training the recommendation model through performing:
- sequentially updating model parameters corresponding to features with frequencies greater than the first threshold in a machine learning model based on the click through rate of a history feature vector of history recommended information in a preset machine learning model, wherein each of the features in the history feature vector of history recommended information in the machine learning model is preset initial values of model parameters corresponding to features, and a frequency of each of the features in the machine learning model is updated using a preset frequency attenuation coefficient upon completion of the training of the machine learning model;
- and
- removing model parameters corresponding to features with frequencies less than or equal to the first threshold upon completion of the training of the machine learning model, and generating the recommendation model based on latest remaining model parameters, wherein the remaining model parameters are model parameters corresponding to the features with the frequencies greater than the first threshold in the machine learning model;
- wherein generating the recommendation model based on the latest remaining model parameters comprises:
- in response to a machine learning model scale reaching a memory limit, filtering model parameters corresponding to features with frequencies less than or equal to a second threshold from the remaining model parameters, and generating the recommendation model based on remaining model parameters acquired upon the filtering, wherein the second threshold is greater than the first threshold.
6 more elements
A recommendation server for recommending information, comprising:
- one or more processors;
- and
- a memory configured to store one or more programs;
- wherein the one or more programs, when loaded and run by the one or more processors, causes the one or more processors to perform a method for recommending information comprising:
- acquiring a click through rate of recommended information by predicting a feature vector of the recommended information by a recommendation model, wherein model parameters corresponding to features that are the same as features of the feature vector and have frequencies less than or equal to a first threshold are removed from the recommendation model;
- and
- issuing the recommended information based on the click through rate, wherein the recommended information is data content corresponding to a current issue request and is looked up by the recommendation server based on a received issue from a client;
- wherein the recommendation model is generated by a system for training the model through performing:
- sequentially updating model parameters corresponding to features with frequencies greater than the first threshold in a machine learning model based on the click through rate of a history feature vector of history recommended information in a preset machine learning model, wherein each of the features in the history feature vector of history recommended information in the machine learning model is preset initial values of model parameters corresponding to features, and a frequency of each of the features in the machine learning model is updated using a preset frequency attenuation coefficient upon completion of the training of the machine learning model;
- and
- removing model parameters corresponding to features with frequencies less than or equal to the first threshold upon completion of the training of the machine learning model, and generating the recommendation model based on latest remaining model parameters, wherein the remaining model parameters are model parameters corresponding to the features with the frequencies greater than the first threshold in the machine learning model;
- wherein generating the recommendation model based on the latest remaining model parameters comprises:
- in response to a machine learning model scale reaching a memory limit, filtering model parameters corresponding to features with frequencies less than or equal to a second threshold from the remaining model parameters, and generating the recommendation model based on remaining model parameters acquired upon the filtering, wherein the second threshold is greater than the first threshold.
10 more elements
Dependencies are inferred from the claim text.