|
I'm familiar (and hence not interested) with co-training and (EM, variational, sampling)-based approaches for generative models. I've seen references on SVM based methods and entropy regularization, but I'm not aware of the standard references for these methods. Is there a general survey on semi-supervised learning, and which are the reference papers on semi-supervised SVMs and entropy regularization? |
|
Have you checked out the SSL survey by Jerry Zhu? It discusses various approaches to SSL including semi-sup SVMs and entropy regularization. Though it doesn't go much deep into entropy regularization but does have some pointers. No, I hadn't. Thanks
(Jul 21 '10 at 21:09)
Alexandre Passos ♦
|
|
An interesting approach is Mann & McCallum's expectation regularization. 1
In case you haven't seen before, check out the posterior regularization (PR) paper that actually shows connections of the above approach and several other related approaches. It also shows that PR in fact leads to a faster optimization as compared to the expectation regularization approach.
(Aug 15 '10 at 21:06)
spinxl39
Thanks! I wasn't aware of it, it seems interesting.
(Aug 15 '10 at 21:09)
Alexandre Passos ♦
|
|
Jerry Zhu has also written a book: Introduction to Semi-Supervised Learning. It is available online, per pay. |
