<?xml version="1.0" encoding="utf-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	>
<channel>
	<title>Comments on: letter recognition network</title>
	<atom:link href="http://verens.com/archives/2007/10/27/letter-recognition-network/feed/" rel="self" type="application/rss+xml" />
	<link>http://verens.com/archives/2007/10/27/letter-recognition-network/</link>
	<description>php, linux, ajax, javascript, kae verens</description>
	<pubDate>Fri, 29 Aug 2008 22:22:33 +0000</pubDate>
	<generator>http://wordpress.org/?v=2.6</generator>
		<item>
		<title>By: Kae Verens</title>
		<link>http://verens.com/archives/2007/10/27/letter-recognition-network/#comment-64828</link>
		<dc:creator>Kae Verens</dc:creator>
		<pubDate>Sun, 28 Oct 2007 10:06:09 +0000</pubDate>
		<guid isPermaLink="false">http://verens.com/archives/2007/10/27/letter-recognition-network/#comment-64828</guid>
		<description>persistence pays off - it now takes only about 150 cycles to learn the entire alphabet, and to prove it, a cycle is run after the training which does not do any adjustments at all to the net while running.</description>
		<content:encoded><![CDATA[<p>persistence pays off - it now takes only about 150 cycles to learn the entire alphabet, and to prove it, a cycle is run after the training which does not do any adjustments at all to the net while running.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: klog &#187; learning how to learn</title>
		<link>http://verens.com/archives/2007/10/27/letter-recognition-network/#comment-64826</link>
		<dc:creator>klog &#187; learning how to learn</dc:creator>
		<pubDate>Sun, 28 Oct 2007 09:38:25 +0000</pubDate>
		<guid isPermaLink="false">http://verens.com/archives/2007/10/27/letter-recognition-network/#comment-64826</guid>
		<description>[...] Yesterday&#8217;s attempts at ANN training seemed at first to be successful, but I had overlooked one simple put curious flaw - there was training going on all the time, and each test was run 10 times&#8230; This means that each neuron was trying different values all the time until it got the right one, then it would be the next neuron&#8217;s turn (a bit of a simplified answer, but I don&#8217;t know how to describe what was actually happening). This ended up causing the tests to look a lot more successful than they actually were. [...]</description>
		<content:encoded><![CDATA[<p>[...] Yesterday&#8217;s attempts at ANN training seemed at first to be successful, but I had overlooked one simple put curious flaw - there was training going on all the time, and each test was run 10 times&#8230; This means that each neuron was trying different values all the time until it got the right one, then it would be the next neuron&#8217;s turn (a bit of a simplified answer, but I don&#8217;t know how to describe what was actually happening). This ended up causing the tests to look a lot more successful than they actually were. [...]</p>
]]></content:encoded>
	</item>
</channel>
</rss>
