<table cellspacing="0" cellpadding="0" border="0" ><tr><td valign="top" style="font: inherit;">Greetings, everyone! I do continue to read this list, even though I don't post, and on the side I continue to work on some theorems related to FBC. The final step in proving one of the theorems requires some linear algebra. I'm hoping somebody can help.<br><br>
<p class="MsoNormal"><a name="OLE_LINK2"></a><a name="OLE_LINK1"><span style="">Suppose we want to define n distinct,
non-overlapping regions in a vector space of some unspecified dimension greater
than or equal to n.<span style=""> </span>The boundaries are
defined by linear equations of the form:</span></a></p>
<p class="MsoNormal"><span style=""><span style=""><o:p> </o:p></span></span></p>
<p class="MsoNormal"><span style=""><span style="">(eq 1)<span style=""> </span>(Nij, v) = 0</span></span></p>
<p class="MsoNormal"><span style=""><span style=""><o:p> </o:p></span></span></p>
<p class="MsoNormal"><span style=""><span style="">Where (Nij, v) is the inner product between a vector v in the space
and the vector Nij normal to the boundary between regions i and j, pointing
from the boundary into region i.<span style=""> </span>(So Nji
= -Nij.)<span style=""> </span>If (Nij,v)>0 then v is on
the i side of the i-j boundary, and if (Nij, v) < 0 then v is on the j side
of the i-j boundary.</span></span></p>
<p class="MsoNormal"><span style=""><span style=""><o:p> </o:p></span></span></p>
<p class="MsoNormal"><span style=""><span style="">The normal vectors Nij and Nik for two different boundaries are
related by symmetric, and orthogonal matrices Sjk satisfying:</span></span></p>
<p class="MsoNormal"><span style=""><span style=""><o:p> </o:p></span></span></p>
<p class="MsoNormal"><span style=""><span style="">(eq 2)<span style=""> </span>Nij = Sjk Nik </span></span></p>
<p class="MsoNormal"><span style=""><span style=""><o:p> </o:p></span></span></p>
<p class="MsoNormal"><span style=""><span style="">(eq 3)<span style=""> </span>Sjk = Skj, Sjk^2 = 1
where 1 is the identity matrix.</span></span></p>
<p class="MsoNormal"><span style=""><span style=""><o:p> </o:p></span></span></p>
<p class="MsoNormal"><span style=""><span style="">We require that every point in the vector space either lie on one of
the boundaries or lie in one of the regions.<span style="">
</span>In general, the conditions (Nij, v) could specify n(n-1)/2 different
surfaces dividing the space into 2^n regions (plus boundaries) but we are
stipulating only n distinct regions.</span></span></p>
<p class="MsoNormal"><span style=""><span style=""><o:p> </o:p></span></span></p>
<p class="MsoNormal"><span style=""><span style="">Does it follow that the normal vectors can ALWAYS be written as:</span></span></p>
<p class="MsoNormal"><span style=""><span style=""><o:p> </o:p></span></span></p>
<p class="MsoNormal"><span style=""><span style="">(eq 4)<span style=""> </span>Nij = Vi-Vj</span></span></p>
<p class="MsoNormal"><span style=""><span style=""><o:p> </o:p></span></span></p>
<p class="MsoNormal"><span style=""><span style="">for all i and j, where Vi and Vj are vectors?</span></span></p>
<p class="MsoNormal"><span style=""><span style=""><o:p> </o:p></span></span></p>
<p class="MsoNormal"><span style=""><span style="">Clearly the vectors defined in eq. 4 satisfy eq. 2.<span style=""> </span>They are not linearly independent, so v
having positive projection onto N12 and N23 means that v also has positive
projection onto N13 = N12 + N23.</span></span></p>
<p class="MsoNormal"><span style=""><span style=""><o:p> </o:p></span></span></p>
<p class="MsoNormal"><span style=""><span style="">So I know that vectors of the form specified in equation 4 satisfy
all of the conditions sketched out above, but I don’t know if the conditions
sketched out above REQUIRE that Nij satisfy 4.</span></span></p></td></tr></table><br>