<div class="gmail_quote">
<div><span lang="ES-AR" style="FONT-SIZE: 9.5pt; FONT-FAMILY: Arial">Seminario de Ecuaciones Diferenciales y Analisis Numerico</span></div>
<div>
<p style="MARGIN: 0cm 0cm 0pt"><span lang="ES-AR" style="FONT-SIZE: 9.5pt; FONT-FAMILY: Arial"> </span></p>
<p style="MARGIN: 0cm 0cm 0pt"><span lang="ES-AR" style="FONT-SIZE: 9.5pt; FONT-FAMILY: Arial">Proxima Charla:</span></p>
<p style="MARGIN: 0cm 0cm 0pt"><span lang="ES-AR" style="FONT-SIZE: 9.5pt; FONT-FAMILY: Arial"> </span></p>
<p style="MARGIN: 0cm 0cm 0pt"><span lang="ES-AR" style="FONT-SIZE: 9.5pt; FONT-FAMILY: Arial"> </span></p>
<p style="MARGIN: 0cm 0cm 0pt"><span lang="ES-AR" style="FONT-SIZE: 9.5pt; FONT-FAMILY: Arial">Martes 1 de Julio 14 hs, Aula E 24- Pab I.</span></p></div>
<div> </div>
<div> </div>
<div>Richard Moore<br><br>New Jersey Inst. of Tech.<br><br>Titulo: Finding Failures in Optical Fibre Lines<br><br>Prior to installation, fibre-optic communication lines must be<br>tested for robustness against environmental sources of noise;<br>
typically, industry standards require that no more than one bit<br>in every billion be lost in transmission. Even where good physical<br>models for transmission exist, they typically involve stochastic<br>nonlinear partial differential equations that render typical<br>
simplifying assumptions (e.g., that the distributions are Gaussian)<br>on transmitted pulse parameters (energy, position, etc.) invalid.<br>Performing simple Monte Carlo simulations on these systems is<br>equally problematic due to the computational expense involved.<br>
<br>We present a method that uses finite-dimensional reductions based<br>on perturbation theory and variational techniques in combination<br>with straightforward sampling methods to build a hybrid<br>analytical-computational approach to resolving these statistically<br>
rare events in optical communication lines. Depending on the<br>quality of the finite-dimensional reduction, the efficiency of this<br>method is several orders of magnitude greater than simple Monte Carlo<br>simulations for a given accuracy. This is joint work with Gino<br>
Biondini (SUNY Buffalo) and Bill Kath (Northwestern Univ.).<br> </div><pre><span lang="ES-AR"><font size="2">Estan todos cordialmente invitados.</font></span></pre><pre><span lang="ES-AR"><font size="2">Sandra Martinez</font></span></pre>
<pre><span lang="ES-AR"><a href="http://mate.dm.uba.ar/~smartin/seminario.htm" target="_blank"><font color="#0000cc" size="2">http://mate.dm.uba.ar/~smartin/seminario.htm</font></a></span></pre></div><br>