Guest Big Time Software Posted June 4, 2000 Share Posted June 4, 2000 Since Lewis thinks (for God knows what warped reason) that I haven't a clue what Fuzzy Logic is, I will explain it It is a mathematical system that is built on the theory that real life situations to not fit into nice, neat, binary (yes/no, good/bad, etc.) categories. Bart Kosko, who is one of the big gurus of Fuzzy Logic, liked to start out his presentations in this way (parapharsing him)... "How many of you are married? (definate number of hands go up) OK, how many of you are happily married? (some hands go down, some stay up where they were, others go down a little, some go up even more)" That is the best way to describe the core concept. As it applies to AI, Fuzzy Logic is focused on the end result instead of the process of obtaining the result. Traditional AI systems are complex, brittle, and CPU intensive. Fuzzy Logic was designed to arrive at the same results, if not better, with a lot less programming, more robustness, and far less computations. It doesn't matter HOW the equations look, but rather what results they produce. Traditional AI is hung up on the systems themselves. And this is why Fuzzy Logic was treated as hearesy by the highly paid AI guys at corporations and learning institutions. My mom was good friends with the head of AI research at GTE and he at least saw the good in it even if he wasn't out preaching it as if it were a new religion The basic concept of the coding is that it is so simple anybody can make a Fuzzy system, even if he/she can't write one line of code. And it works too. I have designed many Fuzzy systems for various games, including a pattern recognition set that is pretty slick if I do say so myself Fuzzy Logic, as I said above, is not based on black and white (binary) thinking. It is instead based on "shades of gray". A Fuzzy value is not 0 or 1, but rather 0 TO 1 (i.e. everything inbetween as well as 0 and 1). So getting back to the marriage example... If you were to rate the Fuzzy degree of marriage happiness of the audience, you first you need a scale: Miserable -> 0 - 1 Unhappy -> 1.1 - 3 Satisified -> 3.1 - 5 Happy -> 5.1 - 7 Very Happy -> 7.1 - 9 Sick in love -> 9.1 - 10 Then you rate each person's response to the question using this scale: Person One -> 0.2 Person Two -> 0.6 Person Three -> 0.1 Person Four -> 0.9 Where the simplicity comes in is something like finding out the average degree of happiness of the marriages. Add up all 4 values and divide by 4. You come up with 4.5. This would mean that the average person was MOSTLY satisified with their marriage. OK, things are a bit more involved than this, but hopefully you get the gist of things Upshot is that you get better and more flexible results for less hits on the CPU and less time spent coding. And CM has Fuzzy Logic all over it, both in the AI and also in the simulation systems. Steve Link to comment Share on other sites More sharing options...
DEF BUNGIS Posted June 4, 2000 Share Posted June 4, 2000 So, if I understand correctly. 1. Fuzzy logic is not a misnomer 2. AI common sense(I am assuming this is the opposite of fuzzy logic), can still be programmed along with fuzzy logic and both running at the same time. (sorry, no nothing about programming) 3. I sense a bit of sarcasim in USERNAME's post. ------------------ Better to make the wrong decision than be the sorry son of a bitch to scared to make one at all Link to comment Share on other sites More sharing options...
Guest Big Time Software Posted June 4, 2000 Share Posted June 4, 2000 Answers... 1. Yes, it is a very well named theory. 2. Yes, traditional AI and Fuzzy Logic systems can run side by side, but it isn't optimal. There is a fusion between some of the traditional AI theories and AI, like Fuzzy Neural Nets. That seems to hold a lot of promise. 3. No, that wasn't sarcasim. Just Lewis doing his usual thing to make himself feel more important. BTW Lewis, Charles *did* weigh in on this when you last stuck your foot in it. But not surprisingly you opted to not respond. Steve Link to comment Share on other sites More sharing options...
:USERNAME: Posted June 5, 2000 Share Posted June 5, 2000 <BLOCKQUOTE>quote:</font><HR>Originally posted by Big Time Software: If you were to rate the Fuzzy degree of marriage happiness of the audience, you first you need a scale: Miserable -> 0 - 1 Unhappy -> 1.1 - 3 Satisified -> 3.1 - 5 Happy -> 5.1 - 7 Very Happy -> 7.1 - 9 Sick in love -> 9.1 - 10 Then you rate each person's response to the question using this scale: Person One -> 0.2 Person Two -> 0.6 Person Three -> 0.1 Person Four -> 0.9 Where the simplicity comes in is something like finding out the average degree of happiness of the marriages. Add up all 4 values and divide by 4. You come up with 4.5. This would mean that the average person was MOSTLY satisified with their marriage. Steve<HR></BLOCKQUOTE> Uh check your math there Steve. I come up with four miserable people. 0.2+0.6+0.1+0.9=1.8 1.8/4=0.45 0.45 is MISERABLE territory. Nothing like someone trying to explain something and screwing the pooch. You ever thought of being a college professor? Lewis Link to comment Share on other sites More sharing options...
Guest Big Time Software Posted June 5, 2000 Share Posted June 5, 2000 Well, at least you stay in character. I mean after making groundless comments about me not understanding Fuzzy Logic it would be against your nature to say "sorry I was an ass" instead of just making an ass out of yourself instead. You are consistant if nothing else. Yeah, one was in whole numbers and the other was in fractions, when both should have been using the same. A readjustment of the ratings: Person One -> 2 Person Two -> 6 Person Three -> 1 Person Four -> 9 Average result of 4.5 I should remind you to not slip fully back into your old ways again. I'm letting the last two comments from you slide, but I am in no mood to take a third from you. Steve [This message has been edited by Big Time Software (edited 06-04-2000).] Link to comment Share on other sites More sharing options...
:USERNAME: Posted June 6, 2000 Share Posted June 6, 2000 <BLOCKQUOTE>quote:</font><HR>Originally posted by Big Time Software: Yeah, one was in whole numbers and the other was in fractions, when both should have been using the same[This message has been edited by Big Time Software (edited 06-04-2000).]<HR></BLOCKQUOTE> But you said that a fuzzy value is between 0 and 1 not 0 or 1. So I am mystifyed by your use of whole numbers. BTW you mean decimals and not fractions (I hope). Lewis Link to comment Share on other sites More sharing options...
tss Posted June 6, 2000 Share Posted June 6, 2000 But you said that a fuzzy value is between 0 and 1 not 0 or 1. So I am mystifyed by your use of whole numbers No need to be mystified. You can always normalize the whole numbers back to the range between 0 and 1 by divining with the largest allowed number. (Actually, your fuzzy engine can do that automatically, and if it doesn't, it really should). For example, suppose that all integers are less than 10. Then you divide with 10 to get the fractions. Me, I'm not in the fuzzy stuff, classical logic is enough for me. (See http://www.tcs.hut.fi/Software/smodels for the logic programming system that I'm currently maintaining). - Tommi Link to comment Share on other sites More sharing options...
Recommended Posts