Dynamic recommendation system using reinforcement learning for continual learning

Abstract

System and methods for predicting content items using a neural network model and performing reinforcement learning as continual learning for training the neural network model includes obtain a first dataset of user actions of a plurality of users at a plurality of user devices and a second dataset of historical data for the plurality of users, extract a first set of embeddings and a second set of embeddings from the first dataset and the second dataset, output a trained model based on applying the first and second set of embeddings, determine a set of candidate content items by the trained model, determine a prediction value for each respective candidate content item of the set of candidate content items by the trained model, and output one or more content items of the set of candidate content items based on the prediction value determined for each respective first candidate content item.