Talk:ADALINE

Perceptron vs ADALINE
I just created the page, but I am having difficulty figuring out the differences between perceptron and ADALINE. Could someone outline the difference please Paskari 19:04, 1 December 2006 (UTC)


 * The difference is only in the training algorithm. Perceprtron training updates depend only on the sign of the weighted sum, relative to the target, while Adaline training uses the value before the sign thresholding.  Dicklyon (talk) 22:57, 5 August 2014 (UTC)

there are broken links in the Delta Learning Rule: ADALINE page, please fix, the images do not display

Merge with Madaline?
User:Neøn proposes to merge with Madaline. Q VVERTYVS (hm?) 06:47, 12 May 2014 (UTC)


 * Comment. Don't be fooled by the name. E.g. a multilayer perceptron is nothing like a perceptron, looking in fact more like a stacked network of logistic regression models. The Madaline page is very incomplete and unclear, but if what it says is true, a Madaline network is probably considerably more involved than Adaline (which is just linear regression as a classifier with SGD training). Q VVERTYVS (hm?) 06:47, 12 May 2014 (UTC)
 * Even if the differences are extensive, if there's only a small amount of content, does it still warrant a separate article? Neøn (talk) 17:39, 13 May 2014 (UTC)
 * I suggest expanding instead of merging; a GScholar search for "Widrow madaline" turns up several papers with hundreds of citations, so their should be enough to write about Madaline alone (and material about Madaline does not necessarily reflect Adaline, since again, the models are very different). Q VVERTYVS (hm?) 07:51, 3 June 2014 (UTC)


 * Support merge – The Adaline and its multi-layer extension are closely enough related that they're best covered in one article. Dicklyon (talk) 22:58, 5 August 2014 (UTC)