It should be possible to write a recursive matrix-oriented orthogonalization. See http://www.jsoftware.com/jwiki/Essays/QR_Decomposition .
On Mon, Sep 15, 2014 at 11:56 PM, 'Bo Jacoby' via Programming < [email protected]> wrote: > I managed to ortogonalize a 10 10 - matrix like this: > > f=.[:(%+./)(]*[:+/[*[)-[*[:+/* > orto =. 3 : 0 > a=.,:{.y > a=.a,{.y=.({:a)f"1}.y > a=.a,{.y=.({:a)f"1}.y > a=.a,{.y=.({:a)f"1}.y > a=.a,{.y=.({:a)f"1}.y > a=.a,{.y=.({:a)f"1}.y > a=.a,{.y=.({:a)f"1}.y > a=.a,{.y=.({:a)f"1}.y > a=.a,{.y=.({:a)f"1}.y > a=.a,{.y=.({:a)f"1}.y > ) > orto(!/~)i.10 > 1 1 1 1 1 1 1 1 1 1 > _9 _7 _5 _3 _1 1 3 5 7 9 > 6 2 _1 _3 _4 _4 _3 _1 2 6 > _42 14 35 31 12 _12 _31 _35 _14 42 > 18 _22 _17 3 18 18 3 _17 _22 18 > _6 14 _1 _11 _6 6 11 1 _14 6 > 3 _11 10 6 _8 _8 6 10 _11 3 > _9 47 _86 42 56 _56 _42 86 _47 9 > 1 _7 20 _28 14 14 _28 20 _7 1 > _1 9 _36 84 _126 126 _84 36 _9 1 > > > How can the program orto be made less embarrassing? > > Thank you! > > > Bo. > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
