<html><head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head><body style="font-family: tt;" text="#000000" bgcolor="#FFFFFF"><div
style="font-family: tt;"><span style="font-family: monospace;">Hi
Mansoureh,<br><br>Yes, there are several ways to solve this problem. If
you have a favourite, you can use Nengo to implement it in various ways
(e.g. with/without spiking, with/without neural dynamics, etc.).<br><br>As
for an NEF specific solution, the circuit described in this MacNeil
paper is an example:<br><br><a class="moz-txt-link-freetext" href="http://compneuro.uwaterloo.ca/publications/macneil2011.html">http://compneuro.uwaterloo.ca/publications/macneil2011.html</a><br><br>In
fact, those methods are more powerful than explored there, as the
feedback error function can be generated by a nonlinear or linear
target, as described more here:<br><br><a class="moz-txt-link-freetext" href="http://compneuro.uwaterloo.ca/publications/bekolay2013.html">http://compneuro.uwaterloo.ca/publications/bekolay2013.html</a><br><br>Just
as a general pointer, if you search for 'learning' here:
<a class="moz-txt-link-freetext" href="http://compneuro.uwaterloo.ca/publications.html">http://compneuro.uwaterloo.ca/publications.html</a> you'll find several
items of potential interest.<br><br>Finally, chapter 9 of the Neural
Engineering book (Eliasmith & Anderson, 2003) has a parameter
estimation example using function representation that could be what
you're after (e.g. if the noise term in your equation changes).<br><br>Best,
.c<br></span><br>Mansoureh Fahimi wrote:<blockquote
cite="mid:CAORPo41hNddWAk09F+2g8-Q7BhFs98xWCnDmaupo+wUkG8SOsQ@mail.gmail.com"
type="cite"><meta http-equiv="Content-Type" content="text/html;
charset=UTF-8"><div dir="ltr">Hello nengo community,<div><br></div><div>I
am trying to implement nengo to perform a simple linear ar regression
of order 1. for this I have an input time series, and I want to find the
parameters c and phi for the following equation: x(t)=c+phi*x(t-1)+e
where e is gaussian noise. There are numerous ways to solve this
problem, but I am curious as to how the Neural Engineering Framework can
be applied to solve this equation for a given time series. </div><div><br></div><div>As
far as I have understood, this is implemented as a simple integrator
plus an unknown constant input (c) in nengo where the transform weight
is also unknown (phi). But I can't figure out how it goes from hereon.
is there a paper with a similar problem you can refer me to?</div><div><br></div><div>I
appreciate your help in advance.</div><div>Sincerely,</div><div><br></div><div>Mansoureh
Fahimi</div></div>
<pre wrap="">_______________________________________________
nengo-user mailing list
<a class="moz-txt-link-abbreviated" href="mailto:nengo-user@ctnsrv.uwaterloo.ca">nengo-user@ctnsrv.uwaterloo.ca</a>
<a class="moz-txt-link-freetext" href="http://ctnsrv.uwaterloo.ca/mailman/listinfo/nengo-user">http://ctnsrv.uwaterloo.ca/mailman/listinfo/nengo-user</a>
</pre></blockquote></div></body></html>