SYSTEM, METHOD, AND COMPUTER PROGRAM PRODUCT FOR RECOMMENDING A SET OF ITEMS TO A USER

Systems, methods, and computer program products obtain training data and train a neural network based on the training data by concatenating a user identifier to each item of a first set of multiple items in a first set of fused embeddings, concatenating the user identifier to each item of a second set of multiple items in a second set of fused embeddings, determining a first score associated with the first set of multiple items based on the first set of fused embeddings, determining a second score associated with the second set of multiple items based on the second set of fused embeddings, and modifying, using an objective function of the neural network that depends on the first score, the second score, and a margin between the first score and the second score, one or more parameters of the neural network