Transfer Learning for Collaborative Filtering Using a Psychometrics Model

Algorithm 2

Transfer learning based on multidimensional psychometrics model (TLMPM).

Input: The data of the source domain denoted by and the training data set of the target domain denoted by .

Output: Parameters , , , and () of the target domain.

() Learn the parameters of the source domain by algorithm CFMPM using , and get parameters () of the

   source domain.

() Set parameters () of the target domain equal to the parameters () of the source domain,

   respectively.

() Initialize parameters , , , , and of the target domain.

() For each iteration

()   Randomly permutate the data in .

()   For each in .

()     Compute by (15);

()     Compute (; ) by (16);

()     Update the parameters according to (17)–(19).

()  End for

()  Decrease , ; they are multiplied by 0.9.

() End for