See also http://www.jsoftware.com/jwiki/Essays/Complete_Tensor#cross_product
----- Original Message ----- From: Ralph G Selfridge <[EMAIL PROTECTED]> Date: Saturday, August 18, 2007 9:17 Subject: Re: [Jprogramming] a vector product To: Programming forum <[email protected]> > As given this is cross product. Some time ago KI had a > discussion about > some sorts of matrix manipulations. The cross product can be > handled as > A+/ .*E +/ .*B > where E is a rank 3 array of 0, 1 and _1 > > When I was working with some symbolic manipulation software this > way of > getting a cross product was faster than anything that involved > indices. > Ralph S > > > On Wed, 15 Aug 2007, Tracy Harms wrote: > > > In "Elementary Matrix Theory", Howard Eves defines a vector > product, for > > three-dimensional (real) Cartesian coordinate use, so: > > > > c = (a2b3 - a3b2, a3b1 - a1b3, a1b2 - a2b1) > > > > (Above, all numerals are subscripts denoting vector elements.) > > > > I implemented that calculation of c with the following definitions: > > > > atomsb =: 1 0 { atomsa =: 3 4 A. i.3 > > vp =: -/@:((atomsa { [) * atomsb { ]) > > > > My question is, does this seem as good a way as any to handle > this? In > > particular, is this a case where it is not worth trying to > shoe-horn the > > functions into a dot phrasing? ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
