Loading ...
Sorry, an error occurred while loading the content.

Re: [EMHL] Re:Line

Expand Messages
  • Nikolaos Dergiades
    Dear Francisco, By the way, If a line L is tangent at P to the circumcircle of ABC and the distances of A, B, C from L are p, q, r do you know if it is
    Message 1 of 19 , Jan 9, 2012
    • 0 Attachment
      Dear Francisco,
      By the way,
      If a line L is tangent at P to the circumcircle of ABC
      and the distances of A, B, C from L are p, q, r
      do you know if it is possible to find the points P
      for which the sum p^2 + q^2 + r^2 becomes min or max?

      Best regards
      Nikos Dergiades


      > Dear Bernard, Martin and Chris, I
      > need to make a correction:
      >
      > The line for which the minimum is attached is always the
      > line
      > (b^2 - c^2) x + (a^2 - b^2 - T) y + (-a^2 + c^2 + T) z == 0
      > and this is always the MAJOR axis of Steiner
      > circum-ellipse.
      >
    • Nikolaos Dergiades
      A remark. If the triangle is equilateral then there is no min or max because p^2 + q^2 + r^2 = 9R^2/2 and obviously p + q + r = 3R ND
      Message 2 of 19 , Jan 10, 2012
      • 0 Attachment
        A remark.
        If the triangle is equilateral then there is no
        min or max because p^2 + q^2 + r^2 = 9R^2/2
        and obviously p + q + r = 3R

        ND

        > Dear Francisco,
        > By the way,
        > If a line L is tangent at P to the circumcircle of ABC
        > and the distances of A, B, C from L are p, q, r
        > do you know if it is possible to find the points P
        > for which the sum p^2 + q^2 + r^2 becomes min or max?
        >
        > Best regards
        > Nikos Dergiades
      • Chris Van Tienhoven
        Dear Francisco, Bernard, Martin and Nikos, Like Francisco stated the major Axis of the Steiner CircumEllipses is the line with the minimal sum of squared
        Message 3 of 19 , Jan 10, 2012
        • 0 Attachment
          Dear Francisco, Bernard, Martin and Nikos,

          Like Francisco stated the major Axis of the Steiner CircumEllipses is the line with the minimal sum of squared distances to the Triangle vertices. Besides that I noticed that of all lines through the Centroid the minor Axis is the line with the maximum sum of squared distances to the Triangle vertices.

          New question:
          When we have 4 points (1:0:0),(0:1:0),(0:0:1),(p:q:r) which line has the minimal sum of squared distances?
          Does it pass through the centroid of the complete quadrangle?

          Best regards,

          Chris van Tienhoven


          --- In Hyacinthos@yahoogroups.com, "Francisco Javier" <garciacapitan@...> wrote:
          >
          > Dear Bernard, Martin and Chris, I need to make a correction:
          >
          > The line for which the minimum is attached is always the line
          > (b^2 - c^2) x + (a^2 - b^2 - T) y + (-a^2 + c^2 + T) z == 0 and this is always the MAJOR axis of Steiner circum-ellipse.
          >
          > --- In Hyacinthos@yahoogroups.com, "Francisco Javier" <garciacapitan@> wrote:
          > >
          > > Dear Bernard, thank you.
          > >
          > > Yes, Coolidge, who cites Vahlen, generalizes for n points, but the minimum is attained by one single line, in our case the minor axis of the Steiner circum-ellipse.
          > >
          > > The minimum sum of squared distances is (1/6) (a^2 + b^2 + c^2 - 2 T) while the value for the major axis is (1/6) (a^2 + b^2 + c^2 + 2 T).
          > >
          > > Best regards,
          > >
          > > Francisco Javier.
          > >
          > > --- In Hyacinthos@yahoogroups.com, Bernard Gibert <bg42@> wrote:
          > > >
          > > > Dear Francisco,
          > > >
          > > > > It's a least squares problem which I believe is well known.
          > > >
          > > >
          > > > a probable reference...
          > > >
          > > > J. L. Coolidge, Two geometrical applications of the method of least squares, MONTHLY 20 (1913) 187�190.
          > > >
          > > > Best regards
          > > >
          > > > Bernard
          > > >
          > > > [Non-text portions of this message have been removed]
          > > >
          > >
          >
        • Chris Van Tienhoven
          Dear friends, ... 1. Thanks Francisco, problem 2 in Coolidge article gives the answer the minimum-square-distances-line passe through the centroid of the
          Message 4 of 19 , Jan 10, 2012
          • 0 Attachment
            Dear friends,

            As an answer to my earlier asked question:
            > When we have 4 points (1:0:0),(0:1:0),(0:0:1),(p:q:r) which line has the minimal sum of squared distances?
            > Does it pass through the centroid of the complete quadrangle?

            1. Thanks Francisco, problem 2 in Coolidge article gives the answer the minimum-square-distances-line passe through the centroid of the complete quadrangle. This is also valid for all sets of n points.
            2. I also found this phrase on 1st page in
            http://www.mamikon.com/USArticles/SumSqrsMonthly.pdf
            "Given a finite set of fixed points in 3-space, what is the locus
            of a point moving in such a way that the sum of the squares of its distances from the fixed points is constant? The answer is both elegant and surprising: the locus is a sphere whose center is at the centroid of the fixed points (if we allow the empty set and a single point as degenerate cases of a sphere)."
            This makes clear that for all sets of n points the sum of square distances is extending in circles (2-D) or spheres (3-D) with the centroid as center. So the Centroid has minimal sum of square distances to n points
            It is not the same as square distances to a line but it gives a rough idea how square distances behave with n points.
            3. In my first calculation of the minimum-square-distances-line in the 4-point case I found also 2 perpendicular lines through the Quadrangle Centroid (similar to the Steiner Axes).

            Best regards,

            Chris van Tienhoven
          • Chris Van Tienhoven
            Dear Friends, The line with minimal sum of squared distances to Triangle vertices is related to the point with minimal sum of squared distances to the sides of
            Message 5 of 19 , Jan 12, 2012
            • 0 Attachment
              Dear Friends,

              The line with minimal sum of squared distances to Triangle vertices is related to the point with minimal sum of squared distances to the sides of a triangle (duality principle).
              Following the instructions from the paper of J.L. Coolidge I figured out which point in a triangle has minimal sum of squared distances to the sides of a reference triangle.
              It is not the X2 (centroid), it is not X13 (Fermat Point).
              It is X6 (Symmedian Point). I wouldn't have expected this.

              Best regards,

              Chris van Tienhoven

              --- In Hyacinthos@yahoogroups.com, "Chris Van Tienhoven" <van10hoven@...> wrote:
              >
              > Dear friends,
              >
              > As an answer to my earlier asked question:
              > > When we have 4 points (1:0:0),(0:1:0),(0:0:1),(p:q:r) which line has the minimal sum of squared distances?
              > > Does it pass through the centroid of the complete quadrangle?
              >
              > 1. Thanks Francisco, problem 2 in Coolidge article gives the answer the minimum-square-distances-line passe through the centroid of the complete quadrangle. This is also valid for all sets of n points.
              > 2. I also found this phrase on 1st page in
              > http://www.mamikon.com/USArticles/SumSqrsMonthly.pdf
              > "Given a finite set of fixed points in 3-space, what is the locus
              > of a point moving in such a way that the sum of the squares of its distances from the fixed points is constant? The answer is both elegant and surprising: the locus is a sphere whose center is at the centroid of the fixed points (if we allow the empty set and a single point as degenerate cases of a sphere)."
              > This makes clear that for all sets of n points the sum of square distances is extending in circles (2-D) or spheres (3-D) with the centroid as center. So the Centroid has minimal sum of square distances to n points
              > It is not the same as square distances to a line but it gives a rough idea how square distances behave with n points.
              > 3. In my first calculation of the minimum-square-distances-line in the 4-point case I found also 2 perpendicular lines through the Quadrangle Centroid (similar to the Steiner Axes).
              >
              > Best regards,
              >
              > Chris van Tienhoven
              >
            • Sung Hyun Lim
              An obvious generalization but, what about point P(r) which has minimal sum of r-th power of distances from P(r) to sides of triangle? P(2)=Symmedian point, as
              Message 6 of 19 , Jan 12, 2012
              • 0 Attachment
                An obvious generalization but,

                what about point P(r) which has minimal sum of r-th power of distances from
                P(r) to sides of triangle? P(2)=Symmedian point, as Chris commented. What
                is the locus for positive r?

                (note that we should take directed distances)

                Regards,
                Sung Hyun


                [Non-text portions of this message have been removed]
              • Barry Wolk
                ... [snip] A symmetric equation of that line is:   x/(-2a^2+b^2+c^2+T) + y/(a^2-2b^2+c^2+T) + z/(a^2+b^2-2c^2+T) = 0 Use the symmetrization technique in
                Message 7 of 19 , Jan 12, 2012
                • 0 Attachment
                  > From: Francisco Javier <garciacapitan@...>
                  >
                  > Let's consider this question, although a long time passed since it was
                  > proposed.
                  >
                  > I find by calculations that the desired line goes through the centroid, and the
                  > line given by Martin does; however, I find another solution, with big square
                  > roots, if we use cartesian coordinates.
                  >
                  > With barycentric coordinates, I get a simple although  non-symmetric form of
                  > this line:
                  >
                  > (b^2 - c^2)x + (a^2 - b^2 - T) y + (-a^2 + c^2 + T}z = 0
                  >
                  > where T= sqrt of  (a^4 + b^4 + c^4  - a^2 b^2 - a^2 c^2 - b^2 c^2)
                  >
                  > Best regards,
                  >
                  > Francisco Javier.
                  >
                  >
                  >
                  > --- In Hyacinthos@yahoogroups.com, martin lukarevski <martinton@...>
                  > wrote:
                  >>
                  >> Martin wrote:
                  >>
                  >> >For a given triangle ABC find a line such that the sum
                  >> >AA'^2 + BB'^2 + CC'^2 is minimal where A',B',C'  are
                  >> >orth.projections of A,B,C on the line. 

                  [snip]

                  A symmetric equation of that line is:

                    x/(-2a^2+b^2+c^2+T) + y/(a^2-2b^2+c^2+T) + z/(a^2+b^2-2c^2+T) = 0

                  Use the symmetrization technique in message 19239 on the trilinear pole of your line.
                  --
                  Barry Wolk
                • Nikolaos Dergiades
                  Dear friends, Let L be a tangent to the circumcircle of ABC at the point P and p, q, r be the distances AA , BB , CC of the vertices A, B, C from L. If f(P) =
                  Message 8 of 19 , Jan 12, 2012
                  • 0 Attachment
                    Dear friends,
                    Let L be a tangent to the circumcircle of ABC at the point P
                    and p, q, r be the distances AA', BB', CC' of the vertices A, B, C from L.
                    If f(P) = p^2 + q^2 + r^2 then the problem to find the minimum or maximum
                    of f(P) seems to be a difficult problem. If ABC is isosceles
                    AB = AC = b and BC = a then the problem has the following solution:

                    If the homogeneous barycentric coordinates of P are (x : y : z) where
                    y = a - z and aayz + bbzx + bbxy = 0 and
                    A* = (aa : -2bb : -2bb) is the antipode of A then
                    it is obvious that
                    f(A) = (4bb - aa) / 2
                    and we can find that
                    f(P) = (4bb - aa)(a^4 + b^4 - 4 a^3 z + 6 a^2 z^2 - 4 a z^3 +
                    2 z^4))/(4 (b^2 - a z + z^2)^2)
                    From the derivative of f(P) in z we find that we have extrema when (
                    z - 2a)(2zz - 2az + aa - bb) = 0. (1)
                    If A'B'C', HaHbHc is the medial and orthic triangle of ABC.
                    From A we draw the tangent AK to the Nine point circle of ABC
                    and the circle (A', AK) that meets BC at N, N'
                    The lines AN, AN' meet again the circumcircle at Q, Q'.
                    If NC = z then since AK^2 = AB'.AHb = b.c.cosA / 2 = (bb + cc - aa) / 4
                    we have
                    (NA')^2 = (2bb - aa) / 4 or
                    4( +-(z - a/2) )^2 = 2bb - aa or
                    2zz - 2az + aa - bb = 0 (2) .
                    Since N is the trace on BC
                    of the point Q we conclude from (1) that the extrema of f(P) are
                    given at the points A, A*, Q, Q' and f(Q) = f(Q').
                    From f(P) we can find that f(A*) = (a^4 + 8b^4) / 2(4bb - aa)
                    Eliminating z from (2) and f(P) we get f(Q) = (4bb - aa)(aa + bb) / 2(3bb - aa)
                    We have also
                    f(P) = f(A) - (bb - aa)(4bb - aa)( aa + (2z - b)^2 ) / 4(zz - az + bb)^2 . . . (3)
                    f(P) = f(A*) - (bb - aa)(2z - a)^2 ( (4bb - aa)(2bb - aa) - bb(2z- a)^2 ) / 4(4bb - aa)(zz - az + bb)^2 . . (4)
                    f(P) = f(Q) + (bb - aa)(4bb - aa)(2zz - 2az + aa - bb)^2 / 4(3bb - aa)(zz - az + bb)^2 . . (5)
                    f(A) - f(A*) = 4bb(bb - aa) / (4bb - aa) . . . (6)

                    1) If a <= b or A <= pi / 3 then f(Q) <= f(P) <= f(A) or we have {min, max} at {Q, A}.
                    2) If bb <= aa <= 2bb or pi / 3 <= A <= pi / 2 then f(A) <= f(P) <= f(Q) and {min, max} at {A, Q}
                    3) If 2bb <= aa < 4bb or pi / 2 <= A < pi then there is no Q and from
                    f(A) <= f(P) <= f(A*) we have that {min, max} at {A, A*}.

                    Best regards
                    Nikos Dergiades
                  • Nikolaos Dergiades
                    Sorry for a typo. I my previou message I wrote that the distances p, q, r are AA , BB , CC correct to AA , BB , CC or delete it because it is not used in the
                    Message 9 of 19 , Jan 12, 2012
                    • 0 Attachment
                      Sorry for a typo.
                      I my previou message
                      I wrote that the distances
                      p, q, r are AA', BB', CC'
                      correct to AA", BB", CC"
                      or delete it because it is
                      not used in the sequel
                      and A'B'C' is the medial triangle.
                      ND

                      > Dear friends,
                      > Let L be a tangent to the circumcircle of ABC at the point
                      > P
                      > and p, q, r be the distances AA', BB', CC' of the vertices
                      > A, B, C from L.
                      > If f(P) = p^2 + q^2 + r^2 then the problem  to find
                      > the minimum or maximum
                      > of f(P) seems to be a difficult problem. If ABC is
                      > isosceles
                      > AB = AC = b and BC = a then the problem has the following
                      > solution:
                      >
                      > If the homogeneous barycentric coordinates of P are (x : y
                      > : z) where
                      > y = a - z  and  aayz + bbzx + bbxy = 0  and
                      >
                      > A* = (aa : -2bb : -2bb)  is the antipode of A 
                      > then
                      > it is obvious that
                      > f(A) = (4bb - aa) / 2
                      > and we can find that
                      > f(P) = (4bb - aa)(a^4 + b^4 - 4 a^3 z + 6 a^2 z^2 - 4 a z^3
                      > +
                      >     2 z^4))/(4 (b^2 - a z + z^2)^2)
                      > From the derivative of f(P) in z we find that we have
                      > extrema when (
                      > z - 2a)(2zz - 2az + aa - bb) = 0. (1)
                      > If A'B'C', HaHbHc  is the medial and orthic triangle
                      > of ABC.
                      > From A we draw the tangent AK to the Nine point circle of
                      > ABC
                      > and the circle (A', AK) that meets BC at N, N'
                      > The lines AN, AN' meet again the circumcircle at Q, Q'.
                      > If  NC = z  then since AK^2 = AB'.AHb = b.c.cosA
                      > / 2 = (bb + cc - aa) / 4
                      > we  have
                      > (NA')^2 = (2bb  - aa) / 4    or
                      > 4( +-(z - a/2) )^2 = 2bb - aa   or 
                      > 2zz - 2az + aa - bb = 0  (2) .
                      > Since N is the trace on BC
                      > of the point Q we conclude from (1) that the extrema of
                      > f(P) are
                      > given at the points A, A*, Q, Q' and f(Q) = f(Q').
                      > From f(P) we can find that f(A*) = (a^4 + 8b^4) / 2(4bb -
                      > aa)
                      > Eliminating z from (2) and f(P)  we get f(Q) = (4bb -
                      > aa)(aa + bb) / 2(3bb - aa)
                      > We have  also
                      > f(P) = f(A) - (bb - aa)(4bb - aa)( aa + (2z - b)^2 ) / 4(zz
                      > - az + bb)^2 . . .  (3)
                      > f(P) = f(A*) - (bb - aa)(2z - a)^2 ( (4bb - aa)(2bb -
                      > aa)  - bb(2z- a)^2 ) / 4(4bb - aa)(zz - az + bb)^2 . .
                      > (4)
                      > f(P) = f(Q) + (bb - aa)(4bb - aa)(2zz - 2az + aa - bb)^2 /
                      > 4(3bb - aa)(zz - az + bb)^2  . . (5)
                      > f(A) - f(A*) = 4bb(bb - aa) / (4bb - aa) . . . (6)
                      >
                      > 1) If a <=  b  or  A <= pi / 3 
                      > then f(Q) <= f(P) <= f(A)  or we have {min, max}
                      > at {Q, A}.
                      > 2) If bb <= aa <= 2bb or pi / 3 <= A <= pi / 2
                      > then  f(A) <= f(P) <= f(Q)   and
                      > {min, max} at {A, Q}
                      > 3) If 2bb <= aa < 4bb or pi / 2 <= A < pi 
                      > then there is no Q and from
                      >    f(A) <= f(P) <= f(A*)  we have
                      > that {min, max} at {A, A*}.
                      >
                      > Best regards
                      > Nikos Dergiades
                      >
                      >
                      >
                      >
                      > ------------------------------------
                      >
                      > Yahoo! Groups Links
                      >
                      >
                      >     Hyacinthos-fullfeatured@yahoogroups.com
                      >
                      >
                      >
                    • Francisco Javier García Capitán
                      Nikolaos, Thank you for using and enyoing Baricentricas.nb, and for your solution that I ll read later. note also that there is a function
                      Message 10 of 19 , Jan 12, 2012
                      • 0 Attachment
                        Nikolaos,

                        Thank you for using and enyoing Baricentricas.nb, and for your solution
                        that I'll read later.

                        note also that there is a function CuadradoDistanciaPuntoRecta to calculate
                        the distance from a point to a line.



                        2012/1/13 Nikolaos Dergiades <ndergiades@...>

                        > Sorry for a typo.
                        > I my previou message
                        > I wrote that the distances
                        > p, q, r are AA', BB', CC'
                        > correct to AA", BB", CC"
                        > or delete it because it is
                        > not used in the sequel
                        > and A'B'C' is the medial triangle.
                        > ND
                        >
                        > > Dear friends,
                        > > Let L be a tangent to the circumcircle of ABC at the point
                        > > P
                        > > and p, q, r be the distances AA', BB', CC' of the vertices
                        > > A, B, C from L.
                        > > If f(P) = p^2 + q^2 + r^2 then the problem to find
                        > > the minimum or maximum
                        > > of f(P) seems to be a difficult problem. If ABC is
                        > > isosceles
                        > > AB = AC = b and BC = a then the problem has the following
                        > > solution:
                        > >
                        > > If the homogeneous barycentric coordinates of P are (x : y
                        > > : z) where
                        > > y = a - z and aayz + bbzx + bbxy = 0 and
                        > >
                        > > A* = (aa : -2bb : -2bb) is the antipode of A
                        > > then
                        > > it is obvious that
                        > > f(A) = (4bb - aa) / 2
                        > > and we can find that
                        > > f(P) = (4bb - aa)(a^4 + b^4 - 4 a^3 z + 6 a^2 z^2 - 4 a z^3
                        > > +
                        > > 2 z^4))/(4 (b^2 - a z + z^2)^2)
                        > > From the derivative of f(P) in z we find that we have
                        > > extrema when (
                        > > z - 2a)(2zz - 2az + aa - bb) = 0. (1)
                        > > If A'B'C', HaHbHc is the medial and orthic triangle
                        > > of ABC.
                        > > From A we draw the tangent AK to the Nine point circle of
                        > > ABC
                        > > and the circle (A', AK) that meets BC at N, N'
                        > > The lines AN, AN' meet again the circumcircle at Q, Q'.
                        > > If NC = z then since AK^2 = AB'.AHb = b.c.cosA
                        > > / 2 = (bb + cc - aa) / 4
                        > > we have
                        > > (NA')^2 = (2bb - aa) / 4 or
                        > > 4( +-(z - a/2) )^2 = 2bb - aa or
                        > > 2zz - 2az + aa - bb = 0 (2) .
                        > > Since N is the trace on BC
                        > > of the point Q we conclude from (1) that the extrema of
                        > > f(P) are
                        > > given at the points A, A*, Q, Q' and f(Q) = f(Q').
                        > > From f(P) we can find that f(A*) = (a^4 + 8b^4) / 2(4bb -
                        > > aa)
                        > > Eliminating z from (2) and f(P) we get f(Q) = (4bb -
                        > > aa)(aa + bb) / 2(3bb - aa)
                        > > We have also
                        > > f(P) = f(A) - (bb - aa)(4bb - aa)( aa + (2z - b)^2 ) / 4(zz
                        > > - az + bb)^2 . . . (3)
                        > > f(P) = f(A*) - (bb - aa)(2z - a)^2 ( (4bb - aa)(2bb -
                        > > aa) - bb(2z- a)^2 ) / 4(4bb - aa)(zz - az + bb)^2 . .
                        > > (4)
                        > > f(P) = f(Q) + (bb - aa)(4bb - aa)(2zz - 2az + aa - bb)^2 /
                        > > 4(3bb - aa)(zz - az + bb)^2 . . (5)
                        > > f(A) - f(A*) = 4bb(bb - aa) / (4bb - aa) . . . (6)
                        > >
                        > > 1) If a <= b or A <= pi / 3
                        > > then f(Q) <= f(P) <= f(A) or we have {min, max}
                        > > at {Q, A}.
                        > > 2) If bb <= aa <= 2bb or pi / 3 <= A <= pi / 2
                        > > then f(A) <= f(P) <= f(Q) and
                        > > {min, max} at {A, Q}
                        > > 3) If 2bb <= aa < 4bb or pi / 2 <= A < pi
                        > > then there is no Q and from
                        > > f(A) <= f(P) <= f(A*) we have
                        > > that {min, max} at {A, A*}.
                        > >
                        > > Best regards
                        > > Nikos Dergiades
                        > >
                        > >
                        > >
                        > >
                        > > ------------------------------------
                        > >
                        > > Yahoo! Groups Links
                        > >
                        > >
                        > > Hyacinthos-fullfeatured@yahoogroups.com
                        > >
                        > >
                        > >
                        >
                        >
                        > ------------------------------------
                        >
                        > Yahoo! Groups Links
                        >
                        >
                        >
                        >


                        --
                        ---
                        Francisco Javier García Capitán
                        http://garciacapitan.99on.com


                        [Non-text portions of this message have been removed]
                      • Chris Van Tienhoven
                        Dear friends, So X(6) in a triangle has minimal sum of squared distances to the sides of a reference triangle. I wondered which point Q in a complete
                        Message 11 of 19 , Jan 16, 2012
                        • 0 Attachment
                          Dear friends,

                          So X(6) in a triangle has minimal sum of squared distances to the sides of a reference triangle.
                          I wondered which point Q in a complete quadrilateral has minimal sum of squared distances to the sides of the complete quadrilateral.
                          To calculate this I had to modify the construction method of J.L. Coolidge:
                          Let O (origin), A and B be random non-collinear points.
                          Go = Quadrangle-Centroid of projection points of O on the basic 4 lines of the Reference Quadrilateral.
                          Ga = Quadrangle-Centroid of projection points of O on the 4 lines through point A parallel to the 4 basic lines of the Reference Quadrilateral.
                          Gb = Quadrangle-Centroid of projection points of O on the 4 lines through point B parallel to the 4 basic lines of the Reference Quadrilateral.
                          Let Sa = Ga.Go ^ O.B and Sb = Gb.Go ^ O.A.
                          Construct A1 on line O.A such that Sa.Ga : Ga.Go = O.A : A.A1.
                          Construct B1 on line O.B such that Sb.Gb : Gb.Go = O.B : B.B1.

                          Construct Q such that O.A1.P.B1 is a parallelogram where O and P are opposite vertices. Now Q is the point of Least Squares.

                          Amazing in this method is that starting points O, A, B can be chosen randomly. This makes it possible to calculate the point Q by taking simple values for these points. This has no consequence for the outcome.
                          Calculation when the 4 basic lines of the complete quadrilateral are:
                          L1=(1:0:0), L2=(0:1:0), L3=(0:0:1), L4=(l:m:n),
                          gives 1st coordinate of Q:
                          a^2 (a^2(l-m)(l-n) + b^2(2m-n)(m-l) + c^2(2n-m)(n-l)).
                          The 2nd and 3rd coordinates are determined cyclically.
                          Just a bit more complicated than the 1st coordinate of X(6): a^2.

                          Best regards,

                          Chris van Tienhoven





                          --- In Hyacinthos@yahoogroups.com, "Chris Van Tienhoven" <van10hoven@...> wrote:
                          >
                          > Dear Friends,
                          >
                          > The line with minimal sum of squared distances to Triangle vertices is related to the point with minimal sum of squared distances to the sides of a triangle (duality principle).
                          > Following the instructions from the paper of J.L. Coolidge I figured out which point in a triangle has minimal sum of squared distances to the sides of a reference triangle.
                          > It is not the X2 (centroid), it is not X13 (Fermat Point).
                          > It is X6 (Symmedian Point). I wouldn't have expected this.
                          >
                          > Best regards,
                          >
                          > Chris van Tienhoven
                        • Chris Van Tienhoven
                          Dear Friends, In my last message about the construction of the point with minimal sum of squared distances to the 4 lines of a complete quadrilateral I made a
                          Message 12 of 19 , Jan 25, 2012
                          • 0 Attachment
                            Dear Friends,

                            In my last message about the construction of the point with minimal sum of squared distances to the 4 lines of a complete quadrilateral I made a mistake.
                            I mentioned:
                            Let Sa = Ga.Go ^ O.B and Sb = Gb.Go ^ O.A.
                            This should be:
                            Let Sa = Ga.Go ^ O.Gb and Sb = Gb.Go ^ O.Ga.
                            I am sorry for the inconvenience.
                            best regards,

                            Chris van Tienhoven


                            --- In Hyacinthos@yahoogroups.com, "Chris Van Tienhoven" <van10hoven@...> wrote:
                            >
                            > Dear friends,
                            >
                            > So X(6) in a triangle has minimal sum of squared distances to the sides of a reference triangle.
                            > I wondered which point Q in a complete quadrilateral has minimal sum of squared distances to the sides of the complete quadrilateral.
                            > To calculate this I had to modify the construction method of J.L. Coolidge:
                            > Let O (origin), A and B be random non-collinear points.
                            > Go = Quadrangle-Centroid of projection points of O on the basic 4 lines of the Reference Quadrilateral.
                            > Ga = Quadrangle-Centroid of projection points of O on the 4 lines through point A parallel to the 4 basic lines of the Reference Quadrilateral.
                            > Gb = Quadrangle-Centroid of projection points of O on the 4 lines through point B parallel to the 4 basic lines of the Reference Quadrilateral.
                            > Let Sa = Ga.Go ^ O.B and Sb = Gb.Go ^ O.A.
                            > Construct A1 on line O.A such that Sa.Ga : Ga.Go = O.A : A.A1.
                            > Construct B1 on line O.B such that Sb.Gb : Gb.Go = O.B : B.B1.
                            >
                            > Construct Q such that O.A1.P.B1 is a parallelogram where O and P are opposite vertices. Now Q is the point of Least Squares.
                            >
                            > Amazing in this method is that starting points O, A, B can be chosen randomly. This makes it possible to calculate the point Q by taking simple values for these points. This has no consequence for the outcome.
                            > Calculation when the 4 basic lines of the complete quadrilateral are:
                            > L1=(1:0:0), L2=(0:1:0), L3=(0:0:1), L4=(l:m:n),
                            > gives 1st coordinate of Q:
                            > a^2 (a^2(l-m)(l-n) + b^2(2m-n)(m-l) + c^2(2n-m)(n-l)).
                            > The 2nd and 3rd coordinates are determined cyclically.
                            > Just a bit more complicated than the 1st coordinate of X(6): a^2.
                            >
                            > Best regards,
                            >
                            > Chris van Tienhoven
                          Your message has been successfully submitted and would be delivered to recipients shortly.