<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:wfw="http://wellformedweb.org/CommentAPI/"
	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:slash="http://purl.org/rss/1.0/modules/slash/"
	>

<channel>
	<title>Thesis Neminis &#187; SVM</title>
	<atom:link href="http://thesis.neminis.org/category/svm/feed/" rel="self" type="application/rss+xml" />
	<link>http://thesis.neminis.org</link>
	<description>Diario di lavoro della tesi di Vincenzo Russo / Work-log of Vincenzo Russo’s Thesis</description>
	<lastBuildDate>Mon, 04 Apr 2011 09:06:36 +0000</lastBuildDate>
	<language>en</language>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
	<generator>http://wordpress.org/?v=3.2.1</generator>
		<item>
		<title>Bregman divergences, SVMs and possible implications</title>
		<link>http://thesis.neminis.org/2007/11/06/bregman-divergences-svms-and-possible-implications/</link>
		<comments>http://thesis.neminis.org/2007/11/06/bregman-divergences-svms-and-possible-implications/#comments</comments>
		<pubDate>Tue, 06 Nov 2007 00:26:03 +0000</pubDate>
		<dc:creator>vincenzo russo</dc:creator>
				<category><![CDATA[Bregman]]></category>
		<category><![CDATA[Clustering]]></category>
		<category><![CDATA[Co-clustering]]></category>
		<category><![CDATA[SVC]]></category>
		<category><![CDATA[SVM]]></category>

		<guid isPermaLink="false">http://thesis.neminis.org/2007/11/06/bregman-divergences-svms-and-possible-implications/</guid>
		<description><![CDATA[In order to find a connection between the works studied (Bregman Co-clustering and Support Vector Clustering) we have performed some research. An interesting result are the following paper: The above paper generalizes the Minimum Enclosing Ball (MEB) problem to the &#8230; <a href="http://thesis.neminis.org/2007/11/06/bregman-divergences-svms-and-possible-implications/">Continue reading <span class="meta-nav">&#8594;</span></a>]]></description>
			<content:encoded><![CDATA[<p>In order to find a connection between the works studied (Bregman Co-clustering and Support Vector Clustering) we have performed some research. An interesting result are the following paper:</p>
<ul>
</ul>
<p>The above paper generalizes the Minimum Enclosing Ball (MEB) problem to the Bregman divergences and also provide a generalization of the Bâdoiu-Clarkson (BC) approximation algorith. This is the same algorithm exploited in practical by the Core Vector Machines</p>
<ul>
</ul>
<p>CVMs reformulate the SVMs as a MEB problem. Since they use the BC algorithm and such an algorithm has been generalized to the Bregman divergences, the research on vector machines could have interesting implications.</p>
]]></content:encoded>
			<wfw:commentRss>http://thesis.neminis.org/2007/11/06/bregman-divergences-svms-and-possible-implications/feed/</wfw:commentRss>
		<slash:comments>0</slash:comments>
		</item>
		<item>
		<title>[OT] Star galaxies separation via SVM/CVM classification &#8211; Part 2</title>
		<link>http://thesis.neminis.org/2007/11/04/ot-star-galaxies-separation-via-svmcvm-classification-part-2/</link>
		<comments>http://thesis.neminis.org/2007/11/04/ot-star-galaxies-separation-via-svmcvm-classification-part-2/#comments</comments>
		<pubDate>Sun, 04 Nov 2007 11:14:45 +0000</pubDate>
		<dc:creator>vincenzo russo</dc:creator>
				<category><![CDATA[Astrophysics]]></category>
		<category><![CDATA[Classification]]></category>
		<category><![CDATA[Off-Topic]]></category>
		<category><![CDATA[SVM]]></category>

		<guid isPermaLink="false">http://thesis.neminis.org/2007/11/04/ot-star-galaxies-separation-via-svmcvm-classification-part-2/</guid>
		<description><![CDATA[This is a modification of the experiments in this post. I rapidly built a new training set and this time I use only this training set for training the SVM/CVM. Than, I test the new trained classifier on all three &#8230; <a href="http://thesis.neminis.org/2007/11/04/ot-star-galaxies-separation-via-svmcvm-classification-part-2/">Continue reading <span class="meta-nav">&#8594;</span></a>]]></description>
			<content:encoded><![CDATA[<p>This is a modification of the experiments in <a href="http://thesis.neminis.org/2007/10/14/ot-star-galaxies-separation-via-svm-classification/">this post</a>.</p>
<p>I rapidly built a new training set and this time I use only this training set for training the SVM/CVM. Than, I test the  new trained classifier on all three dataset of the previous post.</p>
<p>The training set contain 500 points and has been built using stars and galaxies from another portion of sky.</p>
<p><strong>New accuracy results (SVM)</strong></p>
<p>Longo 01: 95,96 %<br />
Longo 02: 98,08 %<br />
Longo 03: 97,956 %</p>
<p><strong>New accuracy results (CVM)</strong></p>
<p>Longo 01: 96,31 %<br />
Longo 02: 97,67 %<br />
Longo 03: 97,138 %</p>
<p>Let us consider the Longo 02 tested with CVM. We have</p>
<p>Completeness for Stars: 98,4 %<br />
Contamination for Stars: 4,7 %</p>
<p>Completeness for Galaxies: 95,4 %<br />
Contamination for Galaxies: 1,5 %</p>
]]></content:encoded>
			<wfw:commentRss>http://thesis.neminis.org/2007/11/04/ot-star-galaxies-separation-via-svmcvm-classification-part-2/feed/</wfw:commentRss>
		<slash:comments>0</slash:comments>
		</item>
		<item>
		<title>[OT] Star galaxies separation via SVM/CVM classification</title>
		<link>http://thesis.neminis.org/2007/10/14/ot-star-galaxies-separation-via-svm-classification/</link>
		<comments>http://thesis.neminis.org/2007/10/14/ot-star-galaxies-separation-via-svm-classification/#comments</comments>
		<pubDate>Sun, 14 Oct 2007 16:44:47 +0000</pubDate>
		<dc:creator>vincenzo russo</dc:creator>
				<category><![CDATA[Astrophysics]]></category>
		<category><![CDATA[Classification]]></category>
		<category><![CDATA[Off-Topic]]></category>
		<category><![CDATA[SVM]]></category>

		<guid isPermaLink="false">http://thesis.neminis.org/2007/10/14/ot-star-galaxies-separation-via-svm-classification/</guid>
		<description><![CDATA[We have used some astrophysics star/galaxies datasets for our clustering problems, because they have heavily overlapping clusters. Here we present some results of an SVM classification performed on the same datasets. In fact, S/G separation is usually faced in a &#8230; <a href="http://thesis.neminis.org/2007/10/14/ot-star-galaxies-separation-via-svm-classification/">Continue reading <span class="meta-nav">&#8594;</span></a>]]></description>
			<content:encoded><![CDATA[<p>We have used some astrophysics star/galaxies datasets for our clustering problems, because they have heavily overlapping clusters.</p>
<p>Here we present some results of an SVM classification performed on the same datasets. In fact, S/G separation is usually faced in a supervised way.</p>
<p>We have used a simple nonlinear SVM/<a href="http://www.cse.ust.hk/~ivor/cvm.html">CVM</a> classifier with a linear kernel (K(x,y) = x&#8217; * y).</p>
<p>For each dataset, we have used 5% of it as training set. The rest is the test set.</p>
<p><strong>Datasets: </strong></p>
<p>Longo 01, 2500 items, 2000 stars, 500 galaxies<br />
Longo 02, 9816 items, 2935 stars, 6883 galaxies<br />
Longo 03, 10940 items, 2978 stars, 7964 galaxies</p>
<p><strong>Accuracy results: </strong></p>
<p>Longo 01: 95%<br />
Longo 02: 98,0746%<br />
Longo 03: 97,925%</p>
<p><strong>Accuracy results with <a href="http://www.cse.ust.hk/~ivor/cvm.html">CVM</a>: </strong></p>
<p>Longo 01: 94,98%<br />
Longo 02: 97,5%<br />
Longo 03: 95,2%</p>
<p>Probably, other kernels could lead to better results, but it is necessary to understand in which way tune the hyperparameters, such as the kernel width and the soft margin constant, etc.</p>
]]></content:encoded>
			<wfw:commentRss>http://thesis.neminis.org/2007/10/14/ot-star-galaxies-separation-via-svm-classification/feed/</wfw:commentRss>
		<slash:comments>0</slash:comments>
		</item>
		<item>
		<title>New talk on SVC and MBI Principle</title>
		<link>http://thesis.neminis.org/2007/10/14/new-talk-on-svc-and-mbi-principle/</link>
		<comments>http://thesis.neminis.org/2007/10/14/new-talk-on-svc-and-mbi-principle/#comments</comments>
		<pubDate>Sun, 14 Oct 2007 12:12:15 +0000</pubDate>
		<dc:creator>vincenzo russo</dc:creator>
				<category><![CDATA[Bregman]]></category>
		<category><![CDATA[Clustering]]></category>
		<category><![CDATA[Co-clustering]]></category>
		<category><![CDATA[Kernel Width Estimation]]></category>
		<category><![CDATA[Missing values]]></category>
		<category><![CDATA[SVC]]></category>
		<category><![CDATA[SVM]]></category>

		<guid isPermaLink="false">http://thesis.neminis.org/2007/10/14/new-talk-on-svc-and-mbi-principle/</guid>
		<description><![CDATA[In the Documents section are available the slides entitled: &#8220;Novel Clustering Techniques: Support Vector Methods and Minimum Bregman Information principle&#8221; SVC has been explained with more care because it still is a very experimental technique.]]></description>
			<content:encoded><![CDATA[<p>In the <em><a href="http://thesis.neminis.org/documenti/">Documents</a></em> section <a href="http://thesis.neminis.org/wp-content/plugins/downloads-manager/upload/slide-3.pdf">are available the slides</a> entitled: &#8220;<em>Novel Clustering Techniques: Support Vector Methods and Minimum Bregman Information principle</em>&#8221;</p>
<p>SVC has been explained with more care because it still is a very experimental technique.</p>
]]></content:encoded>
			<wfw:commentRss>http://thesis.neminis.org/2007/10/14/new-talk-on-svc-and-mbi-principle/feed/</wfw:commentRss>
		<slash:comments>0</slash:comments>
		</item>
		<item>
		<title>[OT] Parallel and Distributed SVM</title>
		<link>http://thesis.neminis.org/2007/09/27/ot-parallel-and-distributed-svm/</link>
		<comments>http://thesis.neminis.org/2007/09/27/ot-parallel-and-distributed-svm/#comments</comments>
		<pubDate>Thu, 27 Sep 2007 21:28:36 +0000</pubDate>
		<dc:creator>vincenzo russo</dc:creator>
				<category><![CDATA[Off-Topic]]></category>
		<category><![CDATA[SVM]]></category>

		<guid isPermaLink="false">http://thesis.neminis.org/2007/09/27/ot-parallel-and-distributed-svm/</guid>
		<description><![CDATA[Three slides on Parallel and Distributed SVMs. Download References: http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=1650264 http://esl.inf.cbs.dk/rup/index.php/DSVM http://dm.unife.it/gpdt/ (and links therein) http://www.cs.unibo.it/~roffilli/thesis/TURROPPT.pdf http://www.cs.unibo.it/~roffilli/thesis/TURRON05.pdf http://research.microsoft.com/users/jplatt/smo.html (and links therein)]]></description>
			<content:encoded><![CDATA[<p>Three slides on Parallel and Distributed SVMs.</p>
<p><a href="http://thesis.neminis.org/wp-content/plugins/downloads-manager/upload/svm_distr.pdf">Download</a></p>
<p><strong>References</strong>:</p>
<p>http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=1650264</p>
<p>http://esl.inf.cbs.dk/rup/index.php/DSVM</p>
<p>http://dm.unife.it/gpdt/ (and links therein)</p>
<p>http://www.cs.unibo.it/~roffilli/thesis/TURROPPT.pdf</p>
<p>http://www.cs.unibo.it/~roffilli/thesis/TURRON05.pdf</p>
<p>http://research.microsoft.com/users/jplatt/smo.html (and links therein)</p>
]]></content:encoded>
			<wfw:commentRss>http://thesis.neminis.org/2007/09/27/ot-parallel-and-distributed-svm/feed/</wfw:commentRss>
		<slash:comments>0</slash:comments>
		</item>
		<item>
		<title>SVC: politica per classificazione BSV</title>
		<link>http://thesis.neminis.org/2007/07/10/svc-politica-per-classificazione-bsv/</link>
		<comments>http://thesis.neminis.org/2007/07/10/svc-politica-per-classificazione-bsv/#comments</comments>
		<pubDate>Mon, 09 Jul 2007 23:09:30 +0000</pubDate>
		<dc:creator>vincenzo russo</dc:creator>
				<category><![CDATA[Benchmark]]></category>
		<category><![CDATA[Software]]></category>
		<category><![CDATA[SVC]]></category>
		<category><![CDATA[SVM]]></category>
		<category><![CDATA[Test]]></category>

		<guid isPermaLink="false">http://thesis.neminis.org/2007/07/10/svc-politica-per-classificazione-bsv/</guid>
		<description><![CDATA[L&#8217;algoritmo di Cluster Assignment usato come tutti gli altri proposti in letteratura non tratta esplicitamente la classificaizione dei Bounded Support Vector, ovvero di quei punti che, per effetto del valore della costante di margine morbido, finiscono fuori dalla sfera di &#8230; <a href="http://thesis.neminis.org/2007/07/10/svc-politica-per-classificazione-bsv/">Continue reading <span class="meta-nav">&#8594;</span></a>]]></description>
			<content:encoded><![CDATA[<p>L&#8217;algoritmo di Cluster Assignment usato</p>
<ul>
</ul>
<p>come tutti gli altri proposti in letteratura non tratta esplicitamente la classificaizione dei Bounded Support Vector, ovvero di quei punti che, per effetto del valore della costante di margine morbido, finiscono fuori dalla sfera di descrizione del dominio anche se in realtà fanno parte di una delle classi del problema.</p>
<p>Il Cone Cluster Labeling prevede due passi:</p>
<ul>
<li>classificazione dei SV</li>
<li>classificazione di tutti gli altri punti in relazione ai SV</li>
</ul>
<p>che di fatto comprende anche i BSV in &#8220;tutti gli altri punti&#8221;.</p>
<p>Si è scelto di modificare in questo modo l&#8217;algoritmo:</p>
<ul>
<li>classificazione dei SV</li>
<li>classificazione di tutti gli altri punti (tranne i BSV) in relazione ai SV</li>
<li>classificazione dei BSV in relazione a tutti gli altri punti già classificati</li>
</ul>
<p>Nel caso dell&#8217;IRIS data set, <strong>questa modifica ha portato l&#8217;accuratezza da un valore di 89,333% a un valore del 90%</strong>.</p>
]]></content:encoded>
			<wfw:commentRss>http://thesis.neminis.org/2007/07/10/svc-politica-per-classificazione-bsv/feed/</wfw:commentRss>
		<slash:comments>2</slash:comments>
		</item>
		<item>
		<title>Appunti: differenze tra MLP e SVM</title>
		<link>http://thesis.neminis.org/2007/07/09/mlp-e-svm-differenze/</link>
		<comments>http://thesis.neminis.org/2007/07/09/mlp-e-svm-differenze/#comments</comments>
		<pubDate>Mon, 09 Jul 2007 19:38:49 +0000</pubDate>
		<dc:creator>vincenzo russo</dc:creator>
				<category><![CDATA[Appunti]]></category>
		<category><![CDATA[MLP]]></category>
		<category><![CDATA[SVM]]></category>

		<guid isPermaLink="false">http://thesis.neminis.org/2007/07/09/mlp-e-svm-differenze/</guid>
		<description><![CDATA[I limiti principali del Multi-Layer Perceptron (MLP) sono: la necessità di fissare a priori la struttura della rete, in termini di hidden layers e di numero di neuroni da porre in ognuno di essi l&#8217;eccessiva ampiezza delle maggiorazioni ottenute per &#8230; <a href="http://thesis.neminis.org/2007/07/09/mlp-e-svm-differenze/">Continue reading <span class="meta-nav">&#8594;</span></a>]]></description>
			<content:encoded><![CDATA[<p>I limiti principali del Multi-Layer Perceptron (MLP) sono:</p>
<ul>
<li>la necessità di fissare a priori la struttura della rete, in termini di hidden layers e di numero di neuroni da porre in ognuno di essi</li>
<li>l&#8217;eccessiva ampiezza delle maggiorazioni ottenute per la VC-dimension dei modelli impiegati praticamente</li>
<li>difficoltà di addestramento nel caso di dataset non linearmente separabili:</li>
<ul>
<li>a causa dell&#8217;alto numero di dimensioni dello spazio dei pesi</li>
<li>poiché le tecniche più diffuse, come la back-propagation, permettono di ottenere i pesi della rete risolvendo un problema di ottimizzazione non convesso e non vincolato che, di conseguenza, presenta un numero indeterminato di minimi locali.</li>
</ul>
</ul>
<p>Le SVM superano questi problemi.<br />
Innanzitutto non c&#8217;è la necessità di costruire esplicitamente la funzione non lineare per mappare gli ingressi nello spazio degli attributi. Tramite il kernel trick si opera implicitamente nello spazio degli attributi (equivalente allo spazio degli hidden layers). In questo modo ci si svincola dall&#8217;obbligo di fissare a priori la struttura della rete neurale. Allo stesso tempo si rende le SVM scalabili rispetto a dati di alta dimensionalità.<br />
Inoltre le SVM assicurano una soluzione unica e globale nel caso si scelta un kernel definito positivamente.</p>
]]></content:encoded>
			<wfw:commentRss>http://thesis.neminis.org/2007/07/09/mlp-e-svm-differenze/feed/</wfw:commentRss>
		<slash:comments>0</slash:comments>
		</item>
		<item>
		<title>SVC: prestazioni del cluster assignment</title>
		<link>http://thesis.neminis.org/2007/07/09/prestazioni-del-svc/</link>
		<comments>http://thesis.neminis.org/2007/07/09/prestazioni-del-svc/#comments</comments>
		<pubDate>Mon, 09 Jul 2007 16:22:29 +0000</pubDate>
		<dc:creator>vincenzo russo</dc:creator>
				<category><![CDATA[Benchmark]]></category>
		<category><![CDATA[Cluster Labeling]]></category>
		<category><![CDATA[Clustering]]></category>
		<category><![CDATA[SVC]]></category>
		<category><![CDATA[SVM]]></category>

		<guid isPermaLink="false">http://thesis.neminis.org/2007/07/09/prestazioni-del-svc/</guid>
		<description><![CDATA[With regard of this document, recently we have performed the test described at the paragraph 5 with all of two cluster assignment algorithms implemented: Complete Graph Cluster Labeling (CGCL, the classic one) and the Cone Cluster Labeling, respectively presented in &#8230; <a href="http://thesis.neminis.org/2007/07/09/prestazioni-del-svc/">Continue reading <span class="meta-nav">&#8594;</span></a>]]></description>
			<content:encoded><![CDATA[<p>With regard of <a href="http://thesis.neminis.org/wp-content/plugins/downloads-manager/upload/svc-experiments-results.pdf">this document</a>, recently we have performed the test described at the paragraph 5 with all of two cluster assignment algorithms implemented: Complete Graph Cluster Labeling (CGCL, the classic one) and the Cone Cluster Labeling, respectively presented in</p>
<ul>
</ul>
<p>and</p>
<ul>
</ul>
<p>In the first case we have a total execution time of 280.87 seconds; in the second case only 0.47 seconds was taken. In both cases, 0.25 seconds was taken in the domain description by the SVM. So, we can say that the classic algorithm was slower than CCL about of 99.92%.</p>
<p>The clustering results are the same and they are reported in the document mentioned above.</p>
]]></content:encoded>
			<wfw:commentRss>http://thesis.neminis.org/2007/07/09/prestazioni-del-svc/feed/</wfw:commentRss>
		<slash:comments>0</slash:comments>
		</item>
		<item>
		<title>Support Vector Methods and MBI Principle</title>
		<link>http://thesis.neminis.org/2007/07/06/support-vector-methods-and-mbi-principle/</link>
		<comments>http://thesis.neminis.org/2007/07/06/support-vector-methods-and-mbi-principle/#comments</comments>
		<pubDate>Fri, 06 Jul 2007 14:52:18 +0000</pubDate>
		<dc:creator>vincenzo russo</dc:creator>
				<category><![CDATA[Bregman]]></category>
		<category><![CDATA[Clustering]]></category>
		<category><![CDATA[Co-clustering]]></category>
		<category><![CDATA[Missing values]]></category>
		<category><![CDATA[SVC]]></category>
		<category><![CDATA[SVM]]></category>

		<guid isPermaLink="false">http://thesis.neminis.org/2007/07/06/support-vector-methods-and-mbi-principle/</guid>
		<description><![CDATA[In the Documents section are available the slides entitled: &#8220;Data Clustering: High dimensionality, missing values and noise. Support Vector Methods and Minimum Bregman Information Principle&#8220;]]></description>
			<content:encoded><![CDATA[<p>In the <em><a href="http://thesis.neminis.org/documenti/">Documents</a></em> section <a href="http://thesis.neminis.org/wp-content/plugins/downloads-manager/upload/slide2.pdf">are available the slides</a> entitled: &#8220;<em>Data Clustering: High dimensionality, missing values and noise. Support Vector Methods and Minimum Bregman Information Principle</em>&#8220;</p>
]]></content:encoded>
			<wfw:commentRss>http://thesis.neminis.org/2007/07/06/support-vector-methods-and-mbi-principle/feed/</wfw:commentRss>
		<slash:comments>0</slash:comments>
		</item>
		<item>
		<title>SVC Preliminary Experiments</title>
		<link>http://thesis.neminis.org/2007/07/04/svc-preliminary-experiments/</link>
		<comments>http://thesis.neminis.org/2007/07/04/svc-preliminary-experiments/#comments</comments>
		<pubDate>Wed, 04 Jul 2007 07:42:58 +0000</pubDate>
		<dc:creator>vincenzo russo</dc:creator>
				<category><![CDATA[Benchmark]]></category>
		<category><![CDATA[Clustering]]></category>
		<category><![CDATA[Dataset]]></category>
		<category><![CDATA[Software]]></category>
		<category><![CDATA[SVC]]></category>
		<category><![CDATA[SVM]]></category>
		<category><![CDATA[Test]]></category>

		<guid isPermaLink="false">http://thesis.neminis.org/2007/07/04/svc-preliminary-experiments/</guid>
		<description><![CDATA[In the section Documents is available for download the PDF with the configurations used for tests and related results; is also available the ZIP archive containing the data-sets used for the experiments.]]></description>
			<content:encoded><![CDATA[<p>In the section <a href="http://thesis.neminis.org/documenti/"><em>Documents</em></a> is available for download the <a href="http://thesis.neminis.org/wp-content/plugins/downloads-manager/upload/svc-experiments-results.pdf">PDF</a> with the configurations used for tests and related results; is also available the <a href="http://thesis.neminis.org/wp-content/plugins/downloads-manager/upload/datasets-preliminary-coclustering.zip">ZIP archive</a> containing the data-sets used for the experiments.</p>
]]></content:encoded>
			<wfw:commentRss>http://thesis.neminis.org/2007/07/04/svc-preliminary-experiments/feed/</wfw:commentRss>
		<slash:comments>0</slash:comments>
		</item>
	</channel>
</rss>

