On Wed, Dec 29, 2010 at 2:29 PM, Dmitriy Lyubimov <[email protected]> wrote:
> In their particular case, they considered learnign latent factor model > first and then subsequently freeze its learning but adding another model > that learns based on side information, under inverse link. > The same process was used by the Elkan and Mohan. > They sort of made a case that 1) it doesn't really matter what inverse link > is (they considered Logistic and i think Poisson regressions as examples) > and 2) it doesn't really matter what method (or even model) is used to > train the models under inverse link function. > Likewise with Elkan and Mohan. They used CG and SGD to learn the model and found that freezing latent factors before learning the side factors was helpful. In talking with Charles Elkan, he seemed to think that learning the side factors first would be useful as well. (I may have gotten those reversed)
