| CPC G06Q 30/0631 (2013.01) [H04L 67/535 (2022.05)] | 19 Claims |
|
1. A method for updating a recommendation model, performed by a computing device, the method comprising:
acquiring, by the computing device, sample user data from a user characteristic library stored on a server and sample recommendation object data from a recommendation object characteristic library stored on the server, the sample recommendation object data including data related to sample media objects;
generating model updating sample characteristics according to the sample user data and the sample recommendation object data, the model updating sample characteristics comprising sub-characteristics of at least two characteristic dimensions;
inputting the model updating sample characteristics into the recommendation model to update the recommendation model in real time, wherein the updated recommendation model is configured for performing recommendation on media objects; and
performing local sparsification on model parameters meeting a model sparsification condition in response to determining that the corresponding model parameters of the sub-characteristics in the model updating sample characteristics meet the model sparsification condition in the process of updating the recommendation model,
wherein inputting the model updating sample characteristics to update the recommendation model comprises:
writing the model updating sample characteristics into a message queue, the message queue being of a linear storage structure in a memory of the server;
acquiring, from the message queue, the model updating sample characteristics in real time and writing the model updating sample characteristics into an internal memory queue on the server;
reading the model updating sample characteristics from the internal memory queue in a streaming mode through a reading thread;
distributing the model updating sample characteristics read by the reading thread from the internal memory queue to at least two training threads, the at least two training threads sharing the recommendation model, each training thread being an operation program executed by the processor of the computing device and being configured to perform operation scheduling;
locally updating the model parameters of the recommendation model in each of the training threads according to the received model updating sample characteristics, and setting a lock function for each model parameter to control model parameters updated in different training threads at the same moment to be different;
acquiring the updated recommendation model according to the updated model parameters of each of the training threads; and
monitoring, by the computing device, a time duration of updating the recommendation model, and in response to the time duration being greater than a duration threshold, storing the updated recommendation model in a model library on the server, and importing the updated recommendation model into a model queue,
wherein the method further comprises: recommending, by the computing device, the media objects through the recommendation model, including:
inputting model input characteristics related to candidate media objects and target user data into the recommendation model stored in the model library on the server, to obtain corresponding recommendation probabilities of the model input characteristics;
sorting the candidate media objects corresponding to the candidate recommendation object identifiers according to the corresponding recommendation probabilities of the model input characteristics; and
pushing the sorted candidate media objects to a user terminal corresponding to a target user identifier.
|