<?xml version="1.0" encoding="utf-8"?><rss version="2.0"><channel><title>RSS for Wavelet networks</title><link>http://academic.research.microsoft.com/Rss.aspx?cata=9&amp;id=967093</link><description>Search RSS feed for Microsoft Academic Search</description><generator>MSRA Libra RSS Burner</generator><copyright>(c)2008 Microsoft Corpration, All right reserved.</copyright><pubDate>Wed, 19 Jun 2013 03:48:22 GMT</pubDate><lastBuildDate>Wed, 19 Jun 2013 03:48:22 GMT</lastBuildDate><category /><item><title>Wavelet networks</title><link>http://academic.research.microsoft.com/Publication/967093</link><pubDate>Tue, 18 Jun 2013 20:48:22 GMT</pubDate><guid isPermaLink="false">967093729</guid><description><![CDATA[<div><a href="http://academic.research.microsoft.com/Author/1460020">Q. Zhang</a>, <a href="http://academic.research.microsoft.com/Author/2151415">A. Benveniste</a>:
            
            <span style="margin-left:20px">(Citations:729)</span><span style="margin-left:20px"><a href="http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=165591">view publication</a></span></div><div>A wavelet network concept, which is based on <a href='http://academic.research.microsoft.com/Keyword/73921/wavelet-transform'>wavelet transform</a>  theory, is proposed as an alternative to feedforward neural networks for approximating arbitrary nonlinear functions. The basic idea is to replace the neurons by `wavelons', i.e., computing units obtained by cascading an affine transform and a multidimensional wavelet. Then these affine transforms and the synaptic weights must be identified from possibly noise corrupted input/output data. An algorithm of <a href='http://academic.research.microsoft.com/Keyword/2800/backpropagation'>backpropagation</a>  type is proposed for wavelet network training, and experimental results are reported</div><div></div><div>Journal: <a href="http://academic.research.microsoft.com/Journal/677">IEEE Transactions on Neural Networks</a>, vol. 3, no. 6, pp. 889-898, 1992</div><div />]]></description></item></channel></rss>