GAUSSIAN RANKING USING MATRIX FACTORIZATION

摘要

In one embodiment of the present invention, a training engine teaches a matrix factorization model to rank items for users based on implicit feedback data and a rank loss function. In operation, the training engine approximates a distribution of scores to corresponding ranks as an approximately Gaussian distribution. Based on this distribution, the training engine selects an activation function that smoothly maps between scores and ranks. To train the matrix factorization model, the training engine directly optimizes the rank loss function based on the activation function and implicit feedback data. By contrast, conventional training engines that optimize approximations of the rank loss function are typically less efficient and produce less accurate ranking models.

主权项

1. A computer-implemented method, comprising:
determining an activation function that maps between scores associated with a plurality of items and ranks associated with the plurality of items based on an approximate Gaussian distribution of the scores; computing a first score for input data associated with a first item included in the plurality of items based on a matrix factorization model; computing a first value of a rank loss function based on the first score and the activation function; and modifying one or more elements included in the matrix factorization model based on the first value of the rank loss function.