Running an NMF

User 916 | 11/6/2014, 10:27:07 AM

I'd like to try and perform as straightforward NMF on a user-item dataset and then access the factors for inspection. However, as far as I can tell, there's no obvious way to do this despite NMF being implemented in factorizationrecommender. Can anyone advise whether it is possible to access the factors after using factorizationrecommender.create ?

Comments

User 6 | 11/6/2014, 1:52:09 PM

Take a look at<a href="http://graphlab.com/products/create/docs/graphlab.toolkits.recommender.html"> our documentation</a>:

"A recommender model object can perform key tasks including predict, recommend, evaluate, and save. <b class="Bold">Model attributes and statistics may be obtained via m.get(), where m is a model object. In particular, trained model parameters may be accessed using m.get(‘coefficients’) or equivalently m[‘coefficients’]"</b>

Let us know if you have further questions.