Revision as of 08:38, 23 August 2007 edit Rama (talk \| contribs) Autopatrolled, Extended confirmed users 44,661 edits No edit summary ← Previous edit		Revision as of 14:37, 17 September 2007 edit undo Keretapi~enwiki (talk \| contribs) 3 edits No edit summary Next edit →
Line 3: * Fisher kernels allow PLSA to be used in a discriminative setting, not as a generative model. * Whoever wrote the part about "severe overfitting problems" should provide a reference for that. :* I stumbled upon a paper stating these overfitting problems and added the reference. * In "Evolutions...", _discriminative_ was obviously wrong -- I think what was meant is _generative_ -- that's one way to present LDA. * Added a bullet on the extension to higher-order data

Talk:Probabilistic latent semantic analysis: Difference between revisions