Comments on: Watchword: Online Learning http://hunch.net/?p=277 Machine learning and learning theory research Wed, 08 Sep 2010 11:14:02 -0500 http://wordpress.org/?v=2.8.4 hourly 1 By: Statistics vs. Machine Learning, fight! (also: ML nomenclature's original sin) - Brendan O'Connor's Blog http://hunch.net/?p=277&cpage=1#comment-245227 Statistics vs. Machine Learning, fight! (also: ML nomenclature's original sin) - Brendan O'Connor's Blog Wed, 03 Dec 2008 05:44:47 +0000 http://hunch.net/?p=277#comment-245227 [...] and there is no necessary reason they must be.) And then there are non-traditional settings such as online learning, reinforcement learning, and active learning, where the structure of access to information in play. [...] [...] and there is no necessary reason they must be.) And then there are non-traditional settings such as online learning, reinforcement learning, and active learning, where the structure of access to information in play. [...]

]]> By: Machine Learning (Theory) » Exponentiated Gradient http://hunch.net/?p=277&cpage=1#comment-112768 Machine Learning (Theory) » Exponentiated Gradient Sun, 12 Aug 2007 22:33:34 +0000 http://hunch.net/?p=277#comment-112768 [...] The learning rate c plays a critical role in the theorem, and the best constant setting of c depends on how many total rounds T there are. Tong Zhang likes to think of this algorithm as the stochastic gradient descent with entropy regularization, which makes it clear that when used as an online optimization algorithm, c should be gradually decreased in value. [...] [...] The learning rate c plays a critical role in the theorem, and the best constant setting of c depends on how many total rounds T there are. Tong Zhang likes to think of this algorithm as the stochastic gradient descent with entropy regularization, which makes it clear that when used as an online optimization algorithm, c should be gradually decreased in value. [...]

]]> By: Claire http://hunch.net/?p=277&cpage=1#comment-104609 Claire Mon, 02 Jul 2007 20:43:36 +0000 http://hunch.net/?p=277#comment-104609 Along the lines of 4., I would add: <b>Online Space Constraint</b> An algorithmic constraint that the storage required by the learner does not grow with the number of seen examples. Along the lines of 4., I would add:
Online Space Constraint An algorithmic constraint that the storage required by the learner does not grow with the number of seen examples.

]]>