
<html>

<head>

<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">

<style type="text/css" style="display:none;"><!-- P {margin-top:0;margin-bottom:0;} --></style>

</head>

<body dir="ltr">

<div id="divtagdefaultwrapper" style="font-size:12pt;color:#000000;background-color:#FFFFFF;font-family:Calibri,Arial,Helvetica,sans-serif;">

<p><br>

</p>

<div style="color: rgb(0, 0, 0);">

<div id="divRplyFwdMsg" dir="ltr">

<div></div>

</div>

<div>

<div id="divtagdefaultwrapper" style="font-size:12pt; color:#000000; background-color:#FFFFFF; font-family:Calibri,Arial,Helvetica,sans-serif">

<p>The following technical report is available at <span style="font-family:Helvetica; font-size:9px"><a href="http://hdl.handle.net/1928/31783" id="LPlnk838243" title="http://hdl.handle.net/1928/31783

Cmd+Click or tap to follow the link">http://hdl.handle.net/1928/31783</a> :</span></p>

<p><span style="font-family:Helvetica; font-size:9px"><br>

</span></p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

Title: The Neural Representation of Concepts at the Sensor Level</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

<br>

</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

Author(s):</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

Healy, Michael John</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

Caudell, Thomas Preston</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

<br>

</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

Abstract: </p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

This report presents a mathematical model of the semantics, or</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

meaning, of the connectionist structure and stimulus activity of a neural</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

network, whether artificial or biological. The mathematical model</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

associates concepts about sensed objects with the neuron-like nodes in</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

a neural network and composable concept relationships with the</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

connection pathways in the network. Category-theoretic constructs,</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

specifically colimits, limits, and functors, organize the concept structure</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

and map it to a formal neural network in a structure-preserving manner.</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

Starting with a simple example of a neural vision system, we show that</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

this mathematical model of neural network structure and activity can be</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

used to derive connectionist architectures that work as intended. We</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

also claim an additional advantage of this approach: A properly functioning</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

connectionist architecture has an accompanying concept</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

representation and this representation is both local and distributed.</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

These properties are derived from the category-theoretic formalism</p>

<p></p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

described here.</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

<br>

</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

Sincerely,</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

Mike Healy</p>

<p style="margin-right:0px; margin-left:0px; font-size:9px; line-height:normal; font-family:Helvetica">

<br>

</p>

</div>

</div>

</div>

</div>

</body>

</html>