<?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" xmlns:sy="http://purl.org/rss/1.0/modules/syndication/" xmlns:media="http://search.yahoo.com/mrss/" > <channel> <title> Comments on: Logic: Cold and Fuzzy </title> <atom:link href="https://dangerousintersection.org/2011/02/03/logic-cold-and-fuzzy/feed/" rel="self" type="application/rss+xml" /> <link>https://dangerousintersection.org/2011/02/03/logic-cold-and-fuzzy/?utm_source=rss&utm_medium=rss&utm_campaign=logic-cold-and-fuzzy&utm_source=rss&utm_medium=rss&utm_campaign=logic-cold-and-fuzzy</link> <description>Human Animals at the Crossroads of Science, Religion, Media and Culture</description> <lastBuildDate>Fri, 04 Feb 2011 13:27:05 +0000</lastBuildDate> <sy:updatePeriod> hourly </sy:updatePeriod> <sy:updateFrequency> 1 </sy:updateFrequency> <item> <title> By: Niklaus Pfirsig </title> <link>https://dangerousintersection.org/2011/02/03/logic-cold-and-fuzzy/comment-page-1/#comment-82596</link> <dc:creator><![CDATA[Niklaus Pfirsig]]></dc:creator> <pubDate>Fri, 04 Feb 2011 13:27:05 +0000</pubDate> <guid isPermaLink="false">http://dangerousintersection.org/?p=16326#comment-82596</guid> <description><![CDATA[Several years ago, I worked with a document processing system which employed fuzzy logic to read handwritten application forms. The layman's description of how the system worked provides a straightforward insight into the operation of fuzzy logic systems. In the system, monochrome scans of application forms were input into a program that identified individual character sub-images, then analyzed each sub-image 27 times. Each analysis (processed in parallel in special processes called operators) was biased to favor a certain subset of possible results. For example, operator 1 might be looking for the characters Q, W, E, R, T, 1, or 2 and operator 2 might be looking for O, W, F, B, I, Z, or 7. (Note that there is some overlap). After all operators have reported the letters from their lists with the closest parameters, the results or tallied like votes to calculate a percentage score called the confidence level. Say you have an applicant named Vadim Clio. The first pass yields results among the 27 operators like this value-votes-confidence "V"-8-29.6 "U"-8-29.6 "W"-6-22.2 "J"-2-7.4 "L"-1-3.7 "K"-1-3.7 "Z"-1-3.7 Next the parameters are tuned for the operators and the image is re-analyzed and the new results are: "V"-10-37.0 "W"-8-29.6 "L"-5-18.5 "U"-4-14.8 The third pass results look something like : "V"-18-66.6 "L"-9-33.3 The AI software used a "confidence threshold" that would stop the output the first character with a confidence level above the threshold. our normal threshold level was 70, meaning that the output character would need at least 19 of the 27 votes. In some cases, the AI couldn't resolve the character above the threshold, and after 9 passes, it would tag the character and its containing field for interactive editing. ]]></description> <content:encoded><![CDATA[<p>Several years ago, I worked with a document processing system which employed fuzzy logic to read handwritten application forms.</p> <p> The layman's description of how the system worked provides a straightforward insight into the operation of fuzzy logic systems.</p> <p> In the system, monochrome scans of application forms were input into a program that identified individual character sub-images, then analyzed each sub-image 27 times. Each analysis (processed in parallel in special processes called operators) was biased to favor a certain subset of possible results. For example, operator 1 might be looking for the characters Q, W, E, R, T, 1, or 2 and operator 2 might be looking for O, W, F, B, I, Z, or 7. (Note that there is some overlap).</p> <p> After all operators have reported the letters from their lists with the closest parameters, the results or tallied like votes to calculate a percentage score called the confidence level.</p> <p> Say you have an applicant named Vadim Clio. The first pass yields results among the 27 operators like this</p> <p>value-votes-confidence</p> <p>"V"-8-29.6</p> <p>"U"-8-29.6</p> <p>"W"-6-22.2</p> <p>"J"-2-7.4</p> <p>"L"-1-3.7</p> <p>"K"-1-3.7</p> <p>"Z"-1-3.7</p> <p>Next the parameters are tuned for the operators and the image is re-analyzed and the new results are:</p> <p>"V"-10-37.0</p> <p>"W"-8-29.6</p> <p>"L"-5-18.5</p> <p>"U"-4-14.8</p> <p>The third pass results look something like :</p> <p>"V"-18-66.6</p> <p>"L"-9-33.3</p> <p>The AI software used a "confidence threshold"</p> <p>that would stop the output the first character with a confidence level above the threshold. our normal threshold level was 70, meaning that the output character would need at least 19 of the 27 votes.</p> <p> In some cases, the AI couldn't resolve the character above the threshold, and after 9 passes, it would tag the character and its containing field for interactive editing. </p> ]]></content:encoded> </item> <item> <title> By: Ben </title> <link>https://dangerousintersection.org/2011/02/03/logic-cold-and-fuzzy/comment-page-1/#comment-82562</link> <dc:creator><![CDATA[Ben]]></dc:creator> <pubDate>Thu, 03 Feb 2011 12:43:33 +0000</pubDate> <guid isPermaLink="false">http://dangerousintersection.org/?p=16326#comment-82562</guid> <description><![CDATA[Everything can be expressed in percentages. Well, 99 percent of things can be expressed in percentages. Dan, you might like (or be good at) online texas hold 'em poker. <a href="http://en.wikipedia.org/wiki/Poker_probability_%28Texas_hold_%27em%29" rel="nofollow">http://en.wikipedia.org/wiki/Poker_probability_%2...</a> ]]></description> <content:encoded><![CDATA[<p>Everything can be expressed in percentages. Well, 99 percent of things can be expressed in percentages.</p> <p>Dan, you might like (or be good at) online texas hold 'em poker. </p> <p> <a href="http://en.wikipedia.org/wiki/Poker_probability_%28Texas_hold_%27em%29" rel="nofollow"></a><a href="http://en.wikipedia.org/wiki/Poker_probability_%2" rel="nofollow ugc">http://en.wikipedia.org/wiki/Poker_probability_%2</a>… </p> ]]></content:encoded> </item> </channel> </rss>