<?xml version="1.0" encoding="UTF-8"?>
<Dialog id="" name="" date="" time="" description="" subject="">
        <Body>
            <Topics/>
            <Turn nickname="(#user0#)" genid="5">
                <Utterance genid="6" ref="-1" time="06:13:24" date="09/12/2007" oldid="5" color="" topic="">HI, I will support Naive Bayes clasification technic</Utterance>
                <Utterance genid="7" ref="-1" time="06:15:01" date="09/12/2007" oldid="6" color="" topic="">The Naive Bayes Model is clearly an easy approach to supervised learning of classification tasks...no matter what type are they</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="8">
                <Utterance genid="9" ref="-1" time="06:17:19" date="09/12/2007" oldid="7" color="" topic="">Hi all</Utterance>
                <Utterance genid="10" ref="-1" time="06:17:47" date="09/12/2007" oldid="8" color="" topic="">i'll discuss about hidden markov models</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="11">
                <Utterance genid="12" ref="7" time="06:18:06" date="09/12/2007" oldid="9" color="" topic=""/>
                <Utterance genid="13" ref="-1" time="06:18:14" date="09/12/2007" oldid="10" color="" topic="">hello everybody</Utterance>
                <Utterance genid="14" ref="-1" time="06:19:00" date="09/12/2007" oldid="11" color="" topic="">I'll be supporting the support vector machine methodologies for supervized learning</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="15">
                <Utterance genid="16" ref="-1" time="06:19:33" date="09/12/2007" oldid="12" color="" topic="">Adrian...are you in?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="17">
                <Utterance genid="18" ref="-1" time="06:20:19" date="09/12/2007" oldid="13" color="" topic="">hidden markov models are used cryptanalysis, speech recognition, machine translation and others</Utterance>
                <Utterance genid="19" ref="16" time="06:20:50" date="09/12/2007" oldid="14" color="" topic="">*are used for</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="20">
                <Utterance genid="21" ref="19" time="06:23:10" date="09/12/2007" oldid="15" color="" topic="">support vector machines also have a wide applicability, starting from e-learning to protein structure prediction</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="22">
                <Utterance genid="23" ref="21" time="06:23:46" date="09/12/2007" oldid="16" color="" topic="">What about learning technics?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="24">
                <Utterance genid="25" ref="23" time="06:26:05" date="09/12/2007" oldid="17" color="" topic="">well, for example, in case of e-learning, support vector machine methods could categorize the learning material and present to the student only what he needs to learn</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="26">
                <Utterance genid="27" ref="25" time="06:26:58" date="09/12/2007" oldid="18" color="" topic="">How this work?</Utterance>
                <Utterance genid="28" ref="-1" time="06:28:29" date="09/12/2007" oldid="19" color="" topic="">What is the pricipal key is this solution...and why you support this technic</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="29">
                <Utterance genid="30" ref="25" time="06:28:47" date="09/12/2007" oldid="20" color="" topic="">for example, for each learner we can associate a context, and depending on this context or category, the learner could be presented only those elements that are from his learning category</Utterance>
                <Utterance genid="31" ref="28" time="06:29:31" date="09/12/2007" oldid="21" color="" topic="">I support it because support vector machines scale well to multi-dimensional inputs</Utterance>
                <Utterance genid="32" ref="31" time="06:30:08" date="09/12/2007" oldid="22" color="" topic="">and text, that is associated with e-learning, is likely to be multi-dimensional</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="33">
                <Utterance genid="34" ref="-1" time="06:30:38" date="09/12/2007" oldid="23" color="" topic="">I see...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="35">
                <Utterance genid="36" ref="-1" time="06:31:13" date="09/12/2007" oldid="24" color="" topic=""/>
            </Turn>
            <Turn nickname="(#user0#)" genid="37">
                <Utterance genid="38" ref="-1" time="06:31:48" date="09/12/2007" oldid="25" color="" topic="">From my side: The Naive Bayes Classifier technique is based on the Bayesian theorem and is particularly suited when the dimensionality of the inputs is high.</Utterance>
                <Utterance genid="39" ref="-1" time="06:32:23" date="09/12/2007" oldid="26" color="" topic="">Despite its simplicity, Naive Bayes can often outperform more sophisticated classification methods</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="40">
                <Utterance genid="41" ref="39" time="06:34:06" date="09/12/2007" oldid="27" color="" topic="">support vector machine algorithms are flexible and they can be as simple or as complex as is needed to solve a supervised learning problem</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="42">
                <Utterance genid="43" ref="-1" time="06:34:35" date="09/12/2007" oldid="28" color="" topic="">Naive Bayes is used in: inductive learning, clasification, data mining, clustering etc</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="44">
                <Utterance genid="45" ref="19" time="06:35:04" date="09/12/2007" oldid="29" color="" topic="">can you give us an exemple based Naive Bayes Classifier technique ?</Utterance>
                <Utterance genid="46" ref="43" time="06:35:51" date="09/12/2007" oldid="30" color="" topic="">no metter it reffers to inductive learning, classification, data mining, and so on ...</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="47">
                <Utterance genid="48" ref="-1" time="06:36:36" date="09/12/2007" oldid="31" color="" topic="">k...I will show you an example on the dashboard</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="49">
                <Utterance genid="50" ref="-1" time="06:37:44" date="09/12/2007" oldid="32" color="" topic="">ok</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="51">
                <Utterance genid="52" ref="45" time="06:37:52" date="09/12/2007" oldid="33" color="" topic="">To demonstrate the concept of Naïve Bayes Classification, consider the example displayed in the image from the dashboard</Utterance>
                <Utterance genid="53" ref="-1" time="06:40:01" date="09/12/2007" oldid="34" color="" topic="">We have a simple example: the objects can be classified as either GREEN or RED</Utterance>
                <Utterance genid="54" ref="-1" time="06:40:36" date="09/12/2007" oldid="35" color="" topic="">Let's exclude people with Dalton syndrom..ok?</Utterance>
                <Utterance genid="55" ref="-1" time="06:40:50" date="09/12/2007" oldid="36" color="" topic="">everybody is healthy...ok?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="56">
                <Utterance genid="57" ref="-1" time="06:40:58" date="09/12/2007" oldid="37" color="" topic="">:))</Utterance>
                <Utterance genid="58" ref="-1" time="06:41:00" date="09/12/2007" oldid="38" color="" topic="">good joke</Utterance>
                <Utterance genid="59" ref="-1" time="06:41:29" date="09/12/2007" oldid="39" color="" topic="">as for me, i'm ok</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="60">
                <Utterance genid="61" ref="-1" time="06:41:49" date="09/12/2007" oldid="40" color="" topic="">We will have to decide to which class (RED or GREEN) they belong, based on the currently exiting objects.</Utterance>
                <Utterance genid="62" ref="-1" time="06:42:55" date="09/12/2007" oldid="41" color="" topic="">new arrived object is marked as an empty circle</Utterance>
                <Utterance genid="63" ref="-1" time="06:43:15" date="09/12/2007" oldid="42" color="" topic="">Like in the picture</Utterance>
                <Utterance genid="64" ref="-1" time="06:43:43" date="09/12/2007" oldid="43" color="" topic="">Since there are twice as many GREEN objects as RED, it is reasonable to believe that a new case (which hasn't been observed yet) is twice as likely to have membership GREEN rather than RED.</Utterance>
                <Utterance genid="65" ref="-1" time="06:43:58" date="09/12/2007" oldid="44" color="" topic="">In the Bayesian analysis, this belief is known as the prior probability</Utterance>
                <Utterance genid="66" ref="-1" time="06:44:08" date="09/12/2007" oldid="45" color="" topic="">Prior probabilities are based on previous experience, in this case the percentage of GREEN and RED objects, and often used to predict outcomes before they actually happen</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="67">
                <Utterance genid="68" ref="64" time="06:44:29" date="09/12/2007" oldid="46" color="" topic="">is this criterion the only one for choosing the class of the new object?</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="69">
                <Utterance genid="70" ref="68" time="06:45:14" date="09/12/2007" oldid="47" color="" topic="">no (#user1#)...let me finish</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="71">
                <Utterance genid="72" ref="70" time="06:45:25" date="09/12/2007" oldid="48" color="" topic="">ok\</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="73">
                <Utterance genid="74" ref="-1" time="06:45:40" date="09/12/2007" oldid="49" color="" topic="">i've added a diagram which represents Markov Process</Utterance>
                <Utterance genid="75" ref="-1" time="06:46:33" date="09/12/2007" oldid="50" color="" topic="">so, you can continue (#user0#)</Utterance>
                <Utterance genid="76" ref="-1" time="06:47:03" date="09/12/2007" oldid="51" color="" topic="">sorry to interrupt</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="77">
                <Utterance genid="78" ref="-1" time="06:47:08" date="09/12/2007" oldid="52" color="" topic="">I will .. as soon as I finish uploading an new pic</Utterance>
                <Utterance genid="79" ref="-1" time="06:47:30" date="09/12/2007" oldid="53" color="" topic="">to continue my approch</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="80">
                <Utterance genid="81" ref="76" time="06:48:32" date="09/12/2007" oldid="54" color="" topic="">so, until then, I want to explain you this diagram from the whiteboard</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="82">
                <Utterance genid="83" ref="79" time="06:48:33" date="09/12/2007" oldid="55" color="" topic="">Approach</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="84">
                <Utterance genid="85" ref="-1" time="06:49:41" date="09/12/2007" oldid="56" color="" topic="">it is done, i'll explain after you</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="86">
                <Utterance genid="87" ref="-1" time="06:49:59" date="09/12/2007" oldid="57" color="" topic="">So here is the prior probability that I have mentioned before</Utterance>
                <Utterance genid="88" ref="-1" time="06:50:17" date="09/12/2007" oldid="58" color="" topic="">look here</Utterance>
                <Utterance genid="89" ref="-1" time="06:50:38" date="09/12/2007" oldid="59" color="" topic="">Since there is a total of 60 objects, 40 of which are GREEN and 20 RED, our prior probabilities for class membership are:</Utterance>
                <Utterance genid="90" ref="-1" time="06:51:04" date="09/12/2007" oldid="60" color="" topic="">Prio probability for GREEN = 40/60</Utterance>
                <Utterance genid="91" ref="-1" time="06:51:07" date="09/12/2007" oldid="61" color="" topic="">and</Utterance>
                <Utterance genid="92" ref="-1" time="06:51:23" date="09/12/2007" oldid="62" color="" topic="">Prior probability for RED = 20/60</Utterance>
                <Utterance genid="93" ref="-1" time="06:52:23" date="09/12/2007" oldid="63" color="" topic="">Having formulated our prior probability, we are now ready to classify a new object (WHITE circle)</Utterance>
                <Utterance genid="94" ref="-1" time="06:53:51" date="09/12/2007" oldid="64" color="" topic="">since the objects are well clustered, it is reasonable to assume that the more GREEN (or RED) objects in the vicinity(likehood) of X, the more likely that the new cases belong to that particular color.</Utterance>
                <Utterance genid="95" ref="-1" time="06:54:13" date="09/12/2007" oldid="65" color="" topic="">From this we calculate the likelihood...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="96">
                <Utterance genid="97" ref="-1" time="06:55:29" date="09/12/2007" oldid="66" color="" topic="">aha</Utterance>
                <Utterance genid="98" ref="-1" time="06:55:36" date="09/12/2007" oldid="67" color="" topic="">that's all i think</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="99">
                <Utterance genid="100" ref="95" time="06:55:36" date="09/12/2007" oldid="68" color="" topic="">I will way for a new picture...</Utterance>
                <Utterance genid="101" ref="-1" time="06:55:44" date="09/12/2007" oldid="69" color="" topic="">so, Ionutz...what about HMM?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="102">
                <Utterance genid="103" ref="-1" time="06:55:57" date="09/12/2007" oldid="70" color="" topic="">ok</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="104">
                <Utterance genid="105" ref="95" time="06:56:00" date="09/12/2007" oldid="71" color="" topic="">the classification is computed using geometrical tools?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="106">
                <Utterance genid="107" ref="-1" time="06:56:18" date="09/12/2007" oldid="72" color="" topic="">so. i've uploaded a picture whict represent the markov process</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="108">
                <Utterance genid="109" ref="105" time="06:56:32" date="09/12/2007" oldid="73" color="" topic="">distance between points, etc/</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="110">
                <Utterance genid="111" ref="-1" time="06:56:58" date="09/12/2007" oldid="74" color="" topic="">Each oval shape represents a random variable that can adopt a number of values. The random variable x(t) is the hidden state at time t (with the model from the above diagram, x(e) in {x1, x2, x3}</Utterance>
                <Utterance genid="112" ref="-1" time="06:57:41" date="09/12/2007" oldid="75" color="" topic="">The random variable y(t) is the observation at time t in ( y(t) in {y1, y2, y3} ).</Utterance>
                <Utterance genid="113" ref="-1" time="06:58:03" date="09/12/2007" oldid="76" color="" topic="">The arrows in the diagram (often called a trellis diagram) denote conditional dependencies.</Utterance>
                <Utterance genid="114" ref="-1" time="06:58:17" date="09/12/2007" oldid="77" color="" topic="">From the diagram, it is clear that the value of the hidden variable x(t)</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="115">
                <Utterance genid="116" ref="109" time="06:58:38" date="09/12/2007" oldid="78" color="" topic="">The picture from the dashboard reprezents a clustering process</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="117">
                <Utterance genid="118" ref="-1" time="06:58:43" date="09/12/2007" oldid="79" color="" topic="">(at time t) only depends on the value of the hidden variable x(t − 1) (at time t − 1).</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="119">
                <Utterance genid="120" ref="112" time="06:59:15" date="09/12/2007" oldid="80" color="" topic="">so the hidden markov model is esspecially useful in future estimation problems</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="121">
                <Utterance genid="122" ref="-1" time="06:59:45" date="09/12/2007" oldid="81" color="" topic="">sorry, but i suggent don't overlaps our disscutions</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="123">
                <Utterance genid="124" ref="-1" time="06:59:51" date="09/12/2007" oldid="82" color="" topic="">so...in this point of view clustering uses geometrical distance to determin if the given object is near or far as is class</Utterance>
                <Utterance genid="125" ref="122" time="07:00:04" date="09/12/2007" oldid="83" color="" topic="">sure....Ionutz</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="126">
                <Utterance genid="127" ref="-1" time="07:00:26" date="09/12/2007" oldid="84" color="" topic="">so, if you have more to tell us, i'll expect you to finish</Utterance>
                <Utterance genid="128" ref="-1" time="07:00:45" date="09/12/2007" oldid="85" color="" topic="">:)</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="129">
                <Utterance genid="130" ref="-1" time="07:01:03" date="09/12/2007" oldid="86" color="" topic="">(#user1#)...did SVN uses linear algebra?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="131">
                <Utterance genid="132" ref="-1" time="07:01:05" date="09/12/2007" oldid="87" color="" topic="">otherwise, i'll continue</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="133">
                <Utterance genid="134" ref="132" time="07:02:00" date="09/12/2007" oldid="88" color="" topic="">please go ahead...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="135">
                <Utterance genid="136" ref="130" time="07:02:13" date="09/12/2007" oldid="89" color="" topic=""/>
                <Utterance genid="137" ref="-1" time="07:02:33" date="09/12/2007" oldid="90" color="" topic="">please, Ionut, go ahead</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="138">
                <Utterance genid="139" ref="-1" time="07:02:46" date="09/12/2007" oldid="91" color="" topic="">k</Utterance>
                <Utterance genid="140" ref="-1" time="07:03:27" date="09/12/2007" oldid="92" color="" topic="">so, what i've explained you it's called Markov property</Utterance>
                <Utterance genid="141" ref="-1" time="07:04:13" date="09/12/2007" oldid="93" color="" topic="">but now let's see who we can compute the probability of an observed sequence</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="142">
                <Utterance genid="143" ref="140" time="07:05:13" date="09/12/2007" oldid="94" color="" topic="">I had oportunity to use HMM implementation in an OCR algoritm for my Project Diploma</Utterance>
                <Utterance genid="144" ref="-1" time="07:05:16" date="09/12/2007" oldid="95" color="" topic="">so I'm in theme</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="145">
                <Utterance genid="146" ref="-1" time="07:05:24" date="09/12/2007" oldid="96" color="" topic="">The probability of observing a sequence Y = y(0), y(1), ... y(L-1) of length L is given by n ext formula :</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="147">
                <Utterance genid="148" ref="-1" time="07:05:52" date="09/12/2007" oldid="97" color="" topic="">go ahead ... I'm interested in what you sustain...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="149">
                <Utterance genid="150" ref="-1" time="07:06:18" date="09/12/2007" oldid="98" color="" topic="">P(Y) = SUM ( P(Y|X)P(X) )</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="151">
                <Utterance genid="152" ref="146" time="07:06:33" date="09/12/2007" oldid="99" color="" topic="">what you mean when you say this?...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="153">
                <Utterance genid="154" ref="-1" time="07:06:34" date="09/12/2007" oldid="100" color="" topic="">where where the sum runs over all possible hidden node sequences</Utterance>
                <Utterance genid="155" ref="-1" time="07:07:09" date="09/12/2007" oldid="101" color="" topic="">X = x(0), x(1), x(L-1)</Utterance>
                <Utterance genid="156" ref="-1" time="07:07:30" date="09/12/2007" oldid="102" color="" topic="">when i said what ?</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="157">
                <Utterance genid="158" ref="154" time="07:07:51" date="09/12/2007" oldid="103" color="" topic="">so..there are some hidden X-s and they should be observ by the HMM method?...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="159">
                <Utterance genid="160" ref="-1" time="07:08:29" date="09/12/2007" oldid="104" color="" topic="">exactly</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="161">
                <Utterance genid="162" ref="156" time="07:08:33" date="09/12/2007" oldid="105" color="" topic="">observ -&gt; what it means?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="163">
                <Utterance genid="164" ref="158" time="07:08:42" date="09/12/2007" oldid="106" color="" topic="">I believe that the y's are the hidden ones, is this correct?</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="165">
                <Utterance genid="166" ref="160" time="07:08:42" date="09/12/2007" oldid="107" color="" topic="">k..I got it...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="167">
                <Utterance genid="168" ref="-1" time="07:09:08" date="09/12/2007" oldid="108" color="" topic="">observ means there to determine, to find ...</Utterance>
                <Utterance genid="169" ref="-1" time="07:09:08" date="09/12/2007" oldid="109" color="" topic="">it's a technical word :)</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="170">
                <Utterance genid="171" ref="164" time="07:09:14" date="09/12/2007" oldid="110" color="" topic="">rfering the pic...?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="172">
                <Utterance genid="173" ref="171" time="07:09:24" date="09/12/2007" oldid="111" color="" topic="">yes</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="174">
                <Utterance genid="175" ref="173" time="07:09:43" date="09/12/2007" oldid="112" color="" topic="">ok...I understood</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="176">
                <Utterance genid="177" ref="-1" time="07:10:00" date="09/12/2007" oldid="113" color="" topic="">k</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="178">
                <Utterance genid="179" ref="-1" time="07:10:00" date="09/12/2007" oldid="114" color="" topic="">So, Ionutz....</Utterance>
                <Utterance genid="180" ref="-1" time="07:10:16" date="09/12/2007" oldid="115" color="" topic="">HMM does it needs any training support?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="181">
                <Utterance genid="182" ref="-1" time="07:10:19" date="09/12/2007" oldid="116" color="" topic="">so, le't me give you a concrete example</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="183">
                <Utterance genid="184" ref="182" time="07:10:45" date="09/12/2007" oldid="117" color="" topic="">find my question first</Utterance>
                <Utterance genid="185" ref="-1" time="07:11:15" date="09/12/2007" oldid="118" color="" topic="">X - is the input</Utterance>
                <Utterance genid="186" ref="-1" time="07:11:29" date="09/12/2007" oldid="119" color="" topic="">Y - is the hidden object</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="187">
                <Utterance genid="188" ref="-1" time="07:11:43" date="09/12/2007" oldid="120" color="" topic="">that's what I meant, also</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="189">
                <Utterance genid="190" ref="-1" time="07:12:17" date="09/12/2007" oldid="121" color="" topic="">HMM needs training support, of course</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="191">
                <Utterance genid="192" ref="188" time="07:12:42" date="09/12/2007" oldid="122" color="" topic="">I believe that all supervised learning methods need training support</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="193">
                <Utterance genid="194" ref="190" time="07:12:48" date="09/12/2007" oldid="123" color="" topic="">what about the amount of data for training?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="195">
                <Utterance genid="196" ref="184" time="07:13:29" date="09/12/2007" oldid="124" color="" topic="">i think, but i'm not sure, there are a lot of data for training</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="197">
                <Utterance genid="198" ref="-1" time="07:13:40" date="09/12/2007" oldid="125" color="" topic="">is it big?...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="199">
                <Utterance genid="200" ref="196" time="07:13:44" date="09/12/2007" oldid="126" color="" topic=""/>
                <Utterance genid="201" ref="-1" time="07:14:20" date="09/12/2007" oldid="127" color="" topic="">how about applicability of HMM?</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="202">
                <Utterance genid="203" ref="192" time="07:14:21" date="09/12/2007" oldid="128" color="" topic="">I know...but I want to get to the amount of data used for training...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="204">
                <Utterance genid="205" ref="-1" time="07:14:57" date="09/12/2007" oldid="129" color="" topic="">ok, I missunderstood</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="206">
                <Utterance genid="207" ref="-1" time="07:15:11" date="09/12/2007" oldid="130" color="" topic="">(#user1#), about applicability of HMM i'll give you an example</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="208">
                <Utterance genid="209" ref="-1" time="07:15:20" date="09/12/2007" oldid="131" color="" topic="">ok</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="210">
                <Utterance genid="211" ref="196" time="07:15:40" date="09/12/2007" oldid="132" color="" topic="">ok...one of the advantages of my Naive Bayes technic is that Naive Bayes classifier is that it requires a small amount of training data to estimate the parameters (means and variances of the variables) necessary for classification</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="212">
                <Utterance genid="213" ref="-1" time="07:16:33" date="09/12/2007" oldid="133" color="" topic="">so, the example is ne</Utterance>
                <Utterance genid="214" ref="-1" time="07:16:36" date="09/12/2007" oldid="134" color="" topic="">xt</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="215">
                <Utterance genid="216" ref="211" time="07:16:39" date="09/12/2007" oldid="135" color="" topic="">Because independent variables are assumed, only the variances of the variables for each class need to be determined and not the entire covariance matrix.</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="217">
                <Utterance genid="218" ref="211" time="07:17:15" date="09/12/2007" oldid="136" color="" topic="">regarding training data, the SVM have the disadvantage of requiring more training data than Naive Bayes, but they make up for it</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="219">
                <Utterance genid="220" ref="-1" time="07:17:17" date="09/12/2007" oldid="137" color="" topic="">Assume you have a friend who lives far away and to whom you talk daily over the telephone about what he did that day.</Utterance>
                <Utterance genid="221" ref="-1" time="07:17:47" date="09/12/2007" oldid="138" color="" topic="">Your friend is only interested in three activities: walking in the park, shopping, and cleaning his apartment.</Utterance>
                <Utterance genid="222" ref="-1" time="07:18:24" date="09/12/2007" oldid="139" color="" topic="">The choice of what to do is determined exclusively by the weather on a given day.</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="223">
                <Utterance genid="224" ref="216" time="07:18:35" date="09/12/2007" oldid="140" color="" topic="">isn't the assumption of independent variables also a weakness of Naive Bayes?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="225">
                <Utterance genid="226" ref="-1" time="07:19:05" date="09/12/2007" oldid="141" color="" topic="">You have no definite information about the weather where your friend lives, but you know general trends.</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="227">
                <Utterance genid="228" ref="224" time="07:19:25" date="09/12/2007" oldid="142" color="" topic="">why you said this?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="229">
                <Utterance genid="230" ref="-1" time="07:19:44" date="09/12/2007" oldid="143" color="" topic="">Based on what he tells you he did each day, you try to guess what the weather must have been like.</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="231">
                <Utterance genid="232" ref="228" time="07:19:48" date="09/12/2007" oldid="144" color="" topic="">because the assumption can be sometimes incorrect</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="233">
                <Utterance genid="234" ref="-1" time="07:20:42" date="09/12/2007" oldid="145" color="" topic="">for me..HMM is too difficult to understand..:(</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="235">
                <Utterance genid="236" ref="-1" time="07:20:45" date="09/12/2007" oldid="146" color="" topic="">for current example look to right picture from whiteboard</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="237">
                <Utterance genid="238" ref="234" time="07:21:18" date="09/12/2007" oldid="147" color="" topic=""/>
            </Turn>
            <Turn nickname="(#user2#)" genid="239">
                <Utterance genid="240" ref="-1" time="07:21:34" date="09/12/2007" oldid="148" color="" topic="">In that piece of code, start_probability represents your belief about which state the HMM is in when your friend first calls you (all you know is that it tends to be rainy on average).</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="241">
                <Utterance genid="242" ref="-1" time="07:21:56" date="09/12/2007" oldid="149" color="" topic="">The arameters (X-s) depend on the type of problem which is to be solved..???</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="243">
                <Utterance genid="244" ref="-1" time="07:22:07" date="09/12/2007" oldid="150" color="" topic="">The particular probability distribution used here is not the equilibrium one, which is (given the transition probabilities) actually approximately {'Rainy': 0.571, 'Sunny': 0.429}.</Utterance>
                <Utterance genid="245" ref="-1" time="07:22:59" date="09/12/2007" oldid="151" color="" topic="">No, there is a pattern which you can use it for all problem</Utterance>
                <Utterance genid="246" ref="-1" time="07:23:09" date="09/12/2007" oldid="152" color="" topic="">s</Utterance>
                <Utterance genid="247" ref="-1" time="07:23:44" date="09/12/2007" oldid="153" color="" topic="">so, go ahead to example</Utterance>
                <Utterance genid="248" ref="-1" time="07:23:48" date="09/12/2007" oldid="154" color="" topic="">The transition_probability represents the change of the weather in the underlying Markov chain.</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="249">
                <Utterance genid="250" ref="190" time="07:23:53" date="09/12/2007" oldid="155" color="" topic="">DISADVANTAGE of HMM when use training: The algorithm is only as good as your training set but more training is not always good.</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="251">
                <Utterance genid="252" ref="-1" time="07:24:20" date="09/12/2007" oldid="156" color="" topic="">yes, it;s right</Utterance>
                <Utterance genid="253" ref="-1" time="07:24:46" date="09/12/2007" oldid="157" color="" topic="">the algorithm has this disadvantage</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="254">
                <Utterance genid="255" ref="-1" time="07:24:54" date="09/12/2007" oldid="158" color="" topic="">so HMM uses a big amount of data...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="256">
                <Utterance genid="257" ref="-1" time="07:25:01" date="09/12/2007" oldid="159" color="" topic="">as i said you</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="258">
                <Utterance genid="259" ref="-1" time="07:25:05" date="09/12/2007" oldid="160" color="" topic="">mine (Naive Bayes) uses less...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="260">
                <Utterance genid="261" ref="257" time="07:25:10" date="09/12/2007" oldid="161" color="" topic=""/>
            </Turn>
            <Turn nickname="(#user0#)" genid="262">
                <Utterance genid="263" ref="-1" time="07:25:13" date="09/12/2007" oldid="162" color="" topic="">how can you comment this?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="264">
                <Utterance genid="265" ref="263" time="07:25:21" date="09/12/2007" oldid="163" color="" topic=""/>
            </Turn>
            <Turn nickname="(#user2#)" genid="266">
                <Utterance genid="267" ref="-1" time="07:25:52" date="09/12/2007" oldid="164" color="" topic="">your alghoritm is better than mine</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="268">
                <Utterance genid="269" ref="267" time="07:26:48" date="09/12/2007" oldid="165" color="" topic="">I believe that a given methodology is as good as it's application field</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="270">
                <Utterance genid="271" ref="252" time="07:27:02" date="09/12/2007" oldid="166" color="" topic="">and reffering to picture, the emission_probability represents how likely your friend is to perform a certain activity on each day.</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="272">
                <Utterance genid="273" ref="224" time="07:27:27" date="09/12/2007" oldid="167" color="" topic="">(#user1#): ... Naive Bayes method performs best when attribute values approach independence</Utterance>
                <Utterance genid="274" ref="273" time="07:27:51" date="09/12/2007" oldid="168" color="" topic="">For problems where attributes have many complex interactions, there is less reason for optimism..:(</Utterance>
                <Utterance genid="275" ref="274" time="07:28:33" date="09/12/2007" oldid="169" color="" topic="">However, the Naive Bayes Model is a good candidate for a first attempt at learning a new classification task....</Utterance>
                <Utterance genid="276" ref="-1" time="07:29:30" date="09/12/2007" oldid="170" color="" topic="">so...we ca imagine a sistem which uses all thouse 4 technics...with the same goal...</Utterance>
                <Utterance genid="277" ref="-1" time="07:29:56" date="09/12/2007" oldid="171" color="" topic="">and Naive Bayes could &lt;perform first&gt;</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="278">
                <Utterance genid="279" ref="276" time="07:30:07" date="09/12/2007" oldid="172" color="" topic="">but first, let me present you how the support vector machine algorithms work</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="281">
                <Utterance genid="282" ref="-1" time="07:30:56" date="09/12/2007" oldid="174" color="" topic="">first of all, SVM's have a very strong mathematical and statistical foundation</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="283">
                <Utterance genid="284" ref="-1" time="07:30:59" date="09/12/2007" oldid="175" color="" topic="">ok, i think that HMM is more difficult and the candidates are Naive Bayes and SVM</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="285">
                <Utterance genid="286" ref="-1" time="07:31:05" date="09/12/2007" oldid="176" color="" topic="">Hi there Adrian B.</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="288">
                <Utterance genid="289" ref="-1" time="07:31:44" date="09/12/2007" oldid="178" color="" topic="">he left us....</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="291">
                <Utterance genid="292" ref="-1" time="07:32:32" date="09/12/2007" oldid="180" color="" topic="">something is wrong there</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="293">
                <Utterance genid="294" ref="282" time="07:33:12" date="09/12/2007" oldid="181" color="" topic="">and strong math behind means that the results of SVM's are calculated before, not extracted from experiments, and people know how this methods will perform</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="295">
                <Utterance genid="296" ref="-1" time="07:33:39" date="09/12/2007" oldid="182" color="" topic="">i propose take a short break and wait for Adrian</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="297">
                <Utterance genid="298" ref="-1" time="07:33:47" date="09/12/2007" oldid="183" color="" topic="">k</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="299">
                <Utterance genid="300" ref="-1" time="07:34:03" date="09/12/2007" oldid="184" color="" topic="">ok, we'll wait</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="301">
                <Utterance genid="302" ref="-1" time="07:34:10" date="09/12/2007" oldid="185" color="" topic="">if everybody confirm that</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="303">
                <Utterance genid="304" ref="-1" time="07:34:45" date="09/12/2007" oldid="186" color="" topic="">let's wait for Adrian...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="305">
                <Utterance genid="306" ref="-1" time="07:35:35" date="09/12/2007" oldid="187" color="" topic="">meanwhile, is it ok with you if I draw something to help me in my presentation?</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="307">
                <Utterance genid="308" ref="-1" time="07:36:29" date="09/12/2007" oldid="188" color="" topic="">k...this will be a good moment...go on...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="309">
                <Utterance genid="310" ref="-1" time="07:36:30" date="09/12/2007" oldid="189" color="" topic="">I'll take your silence as a positive answer :)</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="311">
                <Utterance genid="312" ref="-1" time="07:40:07" date="09/12/2007" oldid="190" color="" topic="">:))</Utterance>
                <Utterance genid="313" ref="-1" time="07:40:28" date="09/12/2007" oldid="191" color="" topic="">Silence is the Answer</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="314">
                <Utterance genid="315" ref="-1" time="07:40:30" date="09/12/2007" oldid="192" color="" topic="">does this approach use geometrical distance?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="316">
                <Utterance genid="317" ref="-1" time="07:41:01" date="09/12/2007" oldid="193" color="" topic="">it is a presentation that will be easy to understand</Utterance>
                <Utterance genid="318" ref="317" time="07:41:14" date="09/12/2007" oldid="194" color="" topic="">I hope</Utterance>
                <Utterance genid="319" ref="-1" time="07:44:18" date="09/12/2007" oldid="195" color="" topic="">ok, the drawing is done</Utterance>
                <Utterance genid="320" ref="-1" time="07:45:54" date="09/12/2007" oldid="196" color="" topic="">if it's ok with you, I'll present the support vector machines</Utterance>
                <Utterance genid="321" ref="-1" time="07:47:27" date="09/12/2007" oldid="197" color="" topic="">ok, we'll wait for more</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="322">
                <Utterance genid="323" ref="-1" time="07:48:29" date="09/12/2007" oldid="198" color="" topic="">So...I'm totaly agrea that SVN can be used in multiple activity domain...I found an example that in my opinion is very interesting</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="324">
                <Utterance genid="325" ref="323" time="07:49:05" date="09/12/2007" oldid="199" color="" topic="">ok, I'll get to that, but first a quick presentation</Utterance>
                <Utterance genid="326" ref="-1" time="07:49:29" date="09/12/2007" oldid="200" color="" topic="">the SVM's can be used for both classification problems and also regression problems</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="327">
                <Utterance genid="328" ref="326" time="07:50:11" date="09/12/2007" oldid="201" color="" topic="">regrsion you mean when an object changes its state?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="330">
                <Utterance genid="331" ref="326" time="07:50:37" date="09/12/2007" oldid="203" color="" topic="">regression problems are those kind of problems where you have a sampling of a function and the SVM will determine with aproximation that function</Utterance>
                <Utterance genid="332" ref="-1" time="07:51:14" date="09/12/2007" oldid="204" color="" topic="">you said something earlier about geometrical distance</Utterance>
                <Utterance genid="333" ref="332" time="07:51:45" date="09/12/2007" oldid="205" color="" topic="">that is one way to implement the SVM, but not the only one</Utterance>
                <Utterance genid="334" ref="333" time="07:52:01" date="09/12/2007" oldid="206" color="" topic="">because SVM are very flexible</Utterance>
                <Utterance genid="335" ref="-1" time="07:52:13" date="09/12/2007" oldid="207" color="" topic="">let me say what I mean by that</Utterance>
                <Utterance genid="336" ref="335" time="07:52:53" date="09/12/2007" oldid="208" color="" topic="">let's consider a classification problem for which we have the lower left figure</Utterance>
                <Utterance genid="337" ref="336" time="07:53:21" date="09/12/2007" oldid="209" color="" topic="">where there are two categories of objects that the SVM was trained with</Utterance>
                <Utterance genid="338" ref="336" time="07:54:12" date="09/12/2007" oldid="210" color="" topic="">the SVM build a separating hyperplane, in this case, a line, that is the border between those two categories</Utterance>
                <Utterance genid="339" ref="338" time="07:55:25" date="09/12/2007" oldid="211" color="" topic="">this hyperplane is optimal becouse it is chosen as to maximize the distance between the support vectors - the elements that are sorrounded in dotted circles</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="340">
                <Utterance genid="341" ref="338" time="07:55:41" date="09/12/2007" oldid="212" color="" topic="">scuse me for interrupting you...SVN uses fist saome future extraction process?...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="342">
                <Utterance genid="343" ref="341" time="07:56:14" date="09/12/2007" oldid="213" color="" topic="">I don't understand the question</Utterance>
                <Utterance genid="344" ref="-1" time="07:58:05" date="09/12/2007" oldid="214" color="" topic="">any way, I'll continue with the second picture, bottom middle one, where there is presented the way the SVM's get rid of a couple of common machine learning problems : overfitting and underfitting</Utterance>
                <Utterance genid="345" ref="344" time="07:58:26" date="09/12/2007" oldid="215" color="" topic="">underfitting is selecting the separating hyperplane in a very simplistic manner,</Utterance>
                <Utterance genid="346" ref="344" time="07:58:42" date="09/12/2007" oldid="216" color="" topic="">and overfitting is making it too complex</Utterance>
                <Utterance genid="347" ref="344" time="08:00:20" date="09/12/2007" oldid="217" color="" topic="">SVM's handle this by recomputing the input data space by adding a new dimmension to the data that will make it more separated, unlike the second picture, where you see that the categories can't be separated by a simple line</Utterance>
                <Utterance genid="348" ref="347" time="08:01:38" date="09/12/2007" oldid="218" color="" topic="">SVM are flexible because of the so called kernel functions, that map the input data into o space with more dimensions, but where the data can be separated by a hyperplane more easily</Utterance>
                <Utterance genid="349" ref="348" time="08:03:03" date="09/12/2007" oldid="219" color="" topic="">this is also why SVM's don't suffer from the so-called dimensionality problem, because they can scale well to multi-dimensional input data</Utterance>
                <Utterance genid="350" ref="349" time="08:05:01" date="09/12/2007" oldid="220" color="" topic="">the applicability of SVM's is very large : facial expression recognition, e-learning, text-classification. wheather and stock exchange forecasting, machine vision, biology, chemestry, etc</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="356">
                <Utterance genid="357" ref="-1" time="10:43:31" date="09/12/2007" oldid="226" color="" topic="">hello guys</Utterance>
                <Utterance genid="358" ref="-1" time="10:43:57" date="09/12/2007" oldid="227" color="" topic="">as you see, i'm back to chat</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="359">
                <Utterance genid="360" ref="-1" time="10:44:35" date="09/12/2007" oldid="228" color="" topic="">is (#user0#) with us?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="361">
                <Utterance genid="362" ref="-1" time="10:44:36" date="09/12/2007" oldid="229" color="" topic="">i propose to continue our discussion</Utterance>
                <Utterance genid="363" ref="-1" time="10:45:29" date="09/12/2007" oldid="230" color="" topic="">yes, he's online</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="364">
                <Utterance genid="365" ref="-1" time="10:46:01" date="09/12/2007" oldid="231" color="" topic="">yes, I'm here...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="366">
                <Utterance genid="367" ref="-1" time="10:46:16" date="09/12/2007" oldid="232" color="" topic="">or better to say connected</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="368">
                <Utterance genid="369" ref="-1" time="10:46:51" date="09/12/2007" oldid="233" color="" topic="">ok guys, I'll let you digest a few seconds what I said in the last part of the conversation</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="371">
                <Utterance genid="372" ref="275" time="10:47:44" date="09/12/2007" oldid="235" color="" topic="">so...we begin talking about a collaboration between thouse technics</Utterance>
                <Utterance genid="373" ref="372" time="10:48:04" date="09/12/2007" oldid="236" color="" topic="">(start talking)</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="374">
                <Utterance genid="375" ref="-1" time="10:48:44" date="09/12/2007" oldid="237" color="" topic="">ok</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="376">
                <Utterance genid="377" ref="-1" time="10:49:15" date="09/12/2007" oldid="238" color="" topic="">We express each of us our pro &amp; contra opinion, but...let's start put them together....and me a good asociation...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="378">
                <Utterance genid="379" ref="372" time="10:49:25" date="09/12/2007" oldid="239" color="" topic="">yes, we can combine these algorithms in order to take advantage of the best feature of each one of them</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="380">
                <Utterance genid="381" ref="-1" time="10:49:32" date="09/12/2007" oldid="240" color="" topic="">we have to choose between SVM and Naive Bayes</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="382">
                <Utterance genid="383" ref="-1" time="10:49:49" date="09/12/2007" oldid="241" color="" topic="">let's start first with Naive Bayes...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="384">
                <Utterance genid="385" ref="381" time="10:50:16" date="09/12/2007" oldid="242" color="" topic="">we can use naive bayes if the input data size is relatively small, and use SVM otherwiae</Utterance>
                <Utterance genid="386" ref="385" time="10:50:25" date="09/12/2007" oldid="243" color="" topic="">*otherwise</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="387">
                <Utterance genid="388" ref="-1" time="10:50:29" date="09/12/2007" oldid="244" color="" topic="">and we conclused HMM is more CPU intensive and difficult than these</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="389">
                <Utterance genid="390" ref="275" time="10:50:41" date="09/12/2007" oldid="245" color="" topic="">yes...true</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="391">
                <Utterance genid="392" ref="-1" time="10:50:51" date="09/12/2007" oldid="246" color="" topic="">i'm sorry for my delay</Utterance>
                <Utterance genid="393" ref="-1" time="10:51:17" date="09/12/2007" oldid="247" color="" topic="">my conection is to slow</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="394">
                <Utterance genid="395" ref="-1" time="10:51:45" date="09/12/2007" oldid="248" color="" topic="">so...we gone have a combination in this order: Naive Bayes -&gt; SVN -&gt; HMM</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="396">
                <Utterance genid="397" ref="-1" time="10:52:00" date="09/12/2007" oldid="249" color="" topic="">too</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="399">
                <Utterance genid="400" ref="-1" time="10:53:18" date="09/12/2007" oldid="251" color="" topic="">let's suppose that we use these technics in a serial mode:...so, every method corresponds to a step in clasification...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="401">
                <Utterance genid="402" ref="395" time="10:53:43" date="09/12/2007" oldid="252" color="" topic="">yes, we can combine the powers of every techniques and come out with a very versatile machine learning software</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="403">
                <Utterance genid="404" ref="-1" time="10:53:45" date="09/12/2007" oldid="253" color="" topic="">and each step can get a result...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="406">
                <Utterance genid="407" ref="404" time="10:55:36" date="09/12/2007" oldid="255" color="" topic="">or we can choose a method depending on the nature of the problem, taking into consideration that each one fits best on a few type of problems</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="408">
                <Utterance genid="409" ref="-1" time="10:56:20" date="09/12/2007" oldid="256" color="" topic="">so...as a conclusion...we can limit to a single method...we can imagine an adaptiv system which could use all of these technics on different situations</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="410">
                <Utterance genid="411" ref="409" time="10:56:39" date="09/12/2007" oldid="257" color="" topic="">yes, that is correct</Utterance>
                <Utterance genid="412" ref="411" time="10:56:49" date="09/12/2007" oldid="258" color="" topic="">after all, no method is perfect</Utterance>
                <Utterance genid="413" ref="412" time="10:57:47" date="09/12/2007" oldid="259" color="" topic=""/>
            </Turn>
            <Turn nickname="(#user0#)" genid="414">
                <Utterance genid="415" ref="-1" time="10:57:56" date="09/12/2007" oldid="260" color="" topic="">so let's shake our hands...for our best...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="416">
                <Utterance genid="417" ref="415" time="10:58:26" date="09/12/2007" oldid="261" color="" topic="">ok, I agree</Utterance>
            </Turn>
            <Turn nickname="(#user3#)" genid="419">
                <Utterance genid="420" ref="279" time="11:10:58" date="09/12/2007" oldid="263" color="" topic=""/>
                <Utterance genid="421" ref="-1" time="11:10:59" date="09/12/2007" oldid="264" color="" topic=""/>
                <Utterance genid="422" ref="-1" time="11:11:00" date="09/12/2007" oldid="265" color="" topic=""/>
                <Utterance genid="423" ref="-1" time="11:11:00" date="09/12/2007" oldid="266" color="" topic=""/>
                <Utterance genid="424" ref="-1" time="11:11:00" date="09/12/2007" oldid="267" color="" topic=""/>
                <Utterance genid="425" ref="-1" time="11:11:03" date="09/12/2007" oldid="268" color="" topic=""/>
                <Utterance genid="426" ref="-1" time="11:11:04" date="09/12/2007" oldid="269" color="" topic=""/>
            </Turn>
        </Body>
    </Dialog>