Cubic spline interpolation with end conditions
pp = csape(x,y)
pp = csape(x,y,conds)
pp = csape(x,[e1,y,e2],conds)
pp = csape({x1,...,xn},___)
returns the cubic spline interpolation to the given data pp
= csape(x
,y
)(x,y)
in ppform
form. The function applies Lagrange end conditions to each end of the data, and matches the
spline endslopes to the slope of the cubic that fits the last four data points at each end.
Data values at the same site are averaged.
returns the cubic spline interpolation for gridded data using the univariate mesh inputs
pp
= csape({x1,...,xn},___)x1,...,xn
. In this case, y
is an
n+r
dimensional array, where r
is the dimensionality
of each data value. conds
is a cell array with n
entries, which provides end conditions for each of the n
variables. In
some cases, you must supply end conditions for end conditions. You can use this syntax with
any of the arguments in the previous syntaxes.
csape
You can implement custom end conditions using the csape
function. Suppose you want to enforce the following condition at the leftmost endpoint, $\mathit{e}=$x(1)
$$\lambda (s)\equiv aDs(e)+b{D}^{2}s(e)=c$$
for the given scalars $\mathit{a}$,$\mathit{b}$, and $\mathit{c}$. You can compute the cubic spine interpolation $\mathit{s}$ as the sum of $${s}_{1}$$ (the cubic spine interpolation of the given data using the default end conditions) and $${s}_{0}$$ (the cubic spine interpolation of zero data using some nontrivial end conditions):
$$s={s}_{1}+\frac{c\lambda ({s}_{1})}{\lambda ({s}_{0})}{s}_{0}$$
The end conditions you specify in $${s}_{0}$$ do not have to be the final desired end conditions $$\lambda (s)$$.
This example uses the titanium test data, a standard data set used in data fitting. Load the data using the titanium
function.
[x,y] = titanium;
Define the coefficients for $$\lambda (s)$$.
a = 2; b = 1; c = 0;
The end condition applies to the leftmost end of the data set.
e = x(1);
Now, calculate the cubic spline interpolation of the data set without imposing the end conditions.
s1 = csape(x,y);
To calculate $${s}_{0}$$, use zero data of the same length as y
with an additional set of nontrivial end conditions.
yZero = zeros(1,length(y));
The 1by2 matrix conds
sets the end conditions by specifying the spline derivatives to fix. This example uses end conditions only on the left end of the data, so use conds
to fix the first derivative at the left end. At the right end, fix the value of the function itself.
conds = [1 0];
To specify the values to fix the function or its derivatives to, add them as additional values to the data set for fitting  in this case, yZero
. The first element specifies the value at the left end, while the last element specifies the value at for the right end.
At the left end, fix the first derivative of the spline to have a value of 1. At the right end, fix the function itself to be 0 (the original value of the final element of yZero
). Concatenate these end condition values at the respecive ends of yZero
and use csape
to find the spline that fits the data with these end condition values.
s0 = csape(x,[1 yZero 0],conds);
Calculate the fully fitted spline from that data by using the aforementioned expression for $\mathit{s}$. To do this, calculate the values for $${\lambda}_{0}=\lambda ({s}_{0})$$ and $${\lambda}_{1}=\lambda ({s}_{1})$$ using the first and second derivatives of the splines $${s}_{0}$$ and $${s}_{1}$$.
d1s1 = fnder(fnbrk(s1,1)); d2s1 = fnder(d1s1); d1s0 = fnder(fnbrk(s0,1)); d2s0 = fnder(d1s0);
Calculate the derivatives of the first polynomial piece of the spline, as the end conditions apply to the left end of the data only.
lam1 = a*fnval(d1s1, e) + b*fnval(d2s1,e); lam0 = a*fnval(d1s0, e) + b*fnval(d2s0,e);
Now use $${\lambda}_{1}$$ and $${\lambda}_{0}$$ to calculate the final, fully fitted spline.
pp = fncmb(s0,(clam1)/lam0,s1);
Plot the spline to compare the results of the default fit and the end conditions.
fnplt(pp,[594, 632]) hold on fnplt(s1,'b',[594, 632]) plot(x,y,'ro','MarkerFaceColor','r') hold off axis([594, 632, 0.62, 0.655]) legend 'Desired end conditions' ... 'Default endconditions' 'Data' ... Location SouthEast
The stationary point near the first data point shows that the end conditions are implemented in the fit.
Use csape
to fit multivariate, vectorvalued data. This example fits vectorvalued data using different end conditions for each independent variable.
First, define the data. For this example, define the 3dimensional vectors v
over a 2dimensional field, with clamped conditions or prescribed slopes in the x
direction and periodic end conditions in the y
direction.
x = 0:4; y = 2:2; s2 = 1/sqrt(2); v = zeros( 3, 7, 5 ); v(1,:,:) = [1 0 s2 1 s2 0 1].'*[1 0 1 0 1]; v(2,:,:) = [1 0 s2 1 s2 0 1].'*[0 1 0 1 0]; v(3,:,:) = [0 1 s2 0 s2 1 0].'*[1 1 1 1 1];
v
is a 3dimensional array with v(:,i+1,j) as
the vector value at coordinate x(i),y(j)
. Two additional entries in the x
dimension specify the slope values: the data points v(:,1,j)
and v(:,7,j)
provide the value of the first derivative along the lines x = 0
and x = 4
for the clamped end conditions. In the y
dimension, the periodic end conditions do not require any additional specification.
Now, calculate the multivariate cubic spline interpolation using csape.
sph = csape({x,y},v,{'clamped','periodic'});
To plot the result, first evaluate the spline over a suitable interval.
values = fnval(sph,{0:.1:4,2:.1:2}); surf(squeeze(values(1,:,:)), ... squeeze(values(2,:,:)), squeeze(values(3,:,:))); axis equal axis off
You can also evaluate and plot the spline surface using the simple command fnplt(sph)
. Note that v
is a 3dimensional array, and v(:,i+1,j)
is the 3vector to match at (x(i),y(j))
, i=1:5
, j=1:5
. Additionally, in accordance with conds{1}
being 'clamped'
, size(v,2)
is 7 (and not 5), and the first and last entry of v(r,:,j)
specify the end slope values.
In some cases, you must supply end conditions of end conditions. In this bivariate example, you reproduce the bicubic polynomial g(x,y) = x^{3}y^{3} by complete bicubic interpolation. You then derive the needed data, including end condition values, directly from g to make it easier to see how the end condition values must be placed. Finally, you check the result.
sites = {[0 1],[0 2]}; coefs = zeros(4, 4); coefs(1,1) = 1; g = ppmak(sites,coefs); Dxg = fnval(fnder(g,[1 0]),sites); Dyg = fnval(fnder(g,[0 1]),sites); Dxyg = fnval(fnder(g,[1 1]),sites); f = csape(sites,[Dxyg(1,1), Dxg(1,:), Dxyg(1,2); ... Dyg(:,1), fnval(g,sites), Dyg(:,2) ; ... Dxyg(2,1), Dxg(2,:), Dxyg(2,2)], ... {'complete','complete'}); if any(squeeze(fnbrk(f,'c'))coefs) disp( 'this is wrong' ) end
x
— Data sitesData sites of data values y
to fit, specified as a vector or as a
cell array for multivariate data. The function creates spline s knots
at each data site x
such that
s(x(j)
) = y(:,j)
for all
j
.
For multivariate, gridded data, specify x
as a cell array that
provides the data site in each variable dimension, such that
s(x1(i),x2(j),...,xn(k)
) =
y(:,i,j,...,k)
.
Data Types: single
 double
y
— Data values to fitData values to fit during creation of the spline, specified as a vector, matrix, or
array. You can specify the data values y(:,j)
as scalars, matrices,
or ndimensional arrays. Data values given at the same data site
x
are averaged.
Data Types: single
 double
conds
— End conditions'clamped'
 'complete'
 'notaknot'
 'periodic'
 'second'
 'variational'
 1by2 matrixEnd conditions for the spline, specified as 'complete'
or
'clamped'
, 'notaknot'
,
'periodic'
, 'second'
,
'variational'
, or as a 1by2 matrix. The predefined options for
conds
impose identical end conditions at each end of the data. You
can specify different end conditions at each end by supplying conds
as a 1by2 matrix.
The available predefined end conditions are as follows.
 Match the endslopes to the given values 
Make the second and secondlast sites inactive knots. This option
ignores any values you provide for  
 Match the first and second derivatives at the left end with those at the right end. 
 Match the end second derivatives to the given values

 Set the end second derivatives equal to zero. This option ignores
any values you provide for 
To specify different end conditions at each end, supply conds
as
a 1by2 matrix. The elements of this matrix elements specify the order of the spline
derivative fixed by the end conditions. Setting conds(j) = i
fixes
the ith derivative D^{i}s
to an end condition value.
The default end condition value is the derivative of the cubic interpolant at the
left four sites when conds(1) = 1
and is 0
otherwise. Set end condition values for the left and right sides of the data by
specifying e1
and e2
, respectively.
You can specify the value of conds(j)
as 0
,
1
, or 2
. If you specify a different value or do
not specify conds(j)
, then conds(j)
is
1
and the corresponding end condition value is the default value.
The following predefined end conditions are available.
clamped  Ds(e) =


curved  D^{2}s(e)
= 

periodic  D^{r}s(a) = D^{r}s(b), r = 1,2 

variational  D^{2}s(e) = 0 

e, a, and b
refer to the left or right data locations; ej
is
e1
for the left end of the data and e2
for the
right end of the data.
You can supply the optional end condition values e1
and
e2
whether you use predefined or userdefined options for
conds
. However, note that some predefined options for
conds
ignore any end condition values you provide.
Example: 'clamped'
, [1 0]
e1
— Left end condition valueLeft end condition value for the spline, specified as a scalar value.
e1
specifies the value for the ith derivative at
the left end of the data, where conds
provides i.
Even if you use different end conditions at the two ends, if you supply an end condition
value at one end you must also supply one for the other end.
Note that some predefined options for conds
ignore any end
condition values you provide.
The default value for e1
is the derivative of the cubic
interpolant at the left four sites when conds(1) = 1
and is
0
otherwise.
Data Types: single
 double
e2
— Right end condition valueRight end condition value for the spline, specified as a scalar value.
e2
specifies the value for the ith derivative at
the right end of the data, where conds
provides i.
Even if you use different end conditions at the two ends, if you supply an end condition
value at one end you must also supply one for the other end.
Note that some predefined options for conds
ignore any end
condition values you provide.
The default value for e2
is the derivative of the cubic
interpolant at the right four sites when conds(2) = 1
and is
0
otherwise.
Data Types: single
 double
pp
— Spline in ppformSpline in ppform, returned as a structure with these fields.
Form
— Form of splinepp
Form of the spline, returned as pp
. pp
indicates that
the spline is given in piecewise polynomial form.
Breaks
— Knot locations of splineKnot positions of the spline, returned as a vector or as a cell array of vectors for multivariate data. Vectors contain strictly increasing elements that represent the start and end of each of the intervals over which the polynomial pieces are defined.
Coefs
— Coefficients of polynomialsCoefficients of polynomials for each piece, returned as a matrix or as an array for multivariate data.
Pieces
— Number of polynomial piecesNumber of polynomial pieces describing the spline, returned as a scalar or as a vector of numbers of pieces in each variable for multivariate data.
Order
— Order of polynomialsOrder of the polynomial function describing each polynomial piece of the spline, returned as a scalar or as a vector containing the order in each variable for multivariate data.
Dim
— DimensionalityDimensionality of the target function, returned as a scalar.
The relevant tridiagonal linear system is constructed and solved using the sparse matrix capabilities of MATLAB^{®}.
The csape
command calls on a much expanded version of the Fortran
routine CUBSPL
in PGS.
You clicked a link that corresponds to this MATLAB command:
Run the command by entering it in the MATLAB Command Window. Web browsers do not support MATLAB commands.
Choose a web site to get translated content where available and see local events and offers. Based on your location, we recommend that you select: .
Select web siteYou can also select a web site from the following list:
Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.