IV
Piecewise Cubic Interpolation;
CUBSPL
Broken lines are neither very smooth nor very efficient approximators.
Both for a smoother approximation and for a more efficient approximation,
one has to go to piecewise polynomial approximation with higher order
pieces. The most popular choice continues to be a cubic approximating
function. In this chapter, we describe various schemes for piecewise cubic
interpolation and give a program for cubic spline interpolation with various
end conditions.
Given the data 9(71),-..,9(tm) with a= 7, <--+ <1» =b, we construct
a piecewise cubic interpolant f to g as follows. On each interval [1;..Ti4:1},
we have f agree with some polynomial P; of order 4,
Q)
f(z) =Pi(z) for 7 [S| 4 40-2) 2, Aald tevin +e).
In matrix form, this reads
2
Ply —a)™D~*(y ~ a) + 5 (1 p)e™ Re
-1Q7,
with D the diagonal matrix [5y1,..-,dyw]. But, by (4), c= 3R-!Q7a, so
that (1) can be written entirely in terms of a as
(8) p(y—a)™D-?(y ~ a) + (1 — p)(R“1Q7 a) R(R*Q7 a).
Since both D-? and (R-1QT)? R(R-!Q7) are (symmetric and) positive
definite, (5) is minimized when a satisfies
~2pD-?(y — a) + 12(1 - p)(R1Q7)*R(R-1Q7)a = 0210 XIV, Smoothing and Lenst-Squares Approximation
or, using the fact that Roh = (ROT and simplifying,
(6) pb Hy - a) = 2(1 = P/Q
In particular,
ss) <9 -9)" 70-9) = | 2=9))'yoett,
compute S(fp) once we know ¢,
with [[b[]3 := 2 [bs]?, which allows us to
we multiply both sides of (6) by
In order to obtain a linear system for ¢,
3QTD? and then use (4) to get
p(3QTy — Re) = (1 — p)Q7D*Qe
or
(6(1 - p)Q7D°Q + pR)c = 3pQ7y.
Now let u be such that
c= 3pu.
Then, in summary,
(7) (6(1 - p)Q7D°Q + pR)u=QTy
and
6) S(fp) = (6(1 ~ p))"1DQuil3,
while, by (6),
(9) a=y-6(1—p)D*Qu.
From thi —
om this, we obtain the information for the ppform of fy 08 follows:
Splzi) = a4
Fplzi) = A |
(0 | a;/ Az, — m1
) Meas a 4 (Sp (2i)/2)Aay ~ (f(a) /6)(Azx4)?
BR) = (e141) Sp (ai) [Ax