SSJ Users Guide

Package functionfit
Function fit utilities
Version: September 29, 2015

This package provides basic facilities for curve fitting and interpolation with polynomials
as, for example, least square fit and spline interpolation.

September 29, 2015

CONTENTS 1

Contents
PolInterp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

LeastSquares . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

BSpline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

SmoothingCubicSpline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

September 29, 2015

PolInterp
Represents a polynomial that interpolates through a set of points. More specifically,
let (x0 , y0 ), . . . , (xn , yn ) be a set of points and p(x) the constructed polynomial of degree n.
Then, for i = 0, . . . , n, p(xi ) = yi .
package umontreal.iro.lecuyer.functionfit;
public class PolInterp extends Polynomial implements Serializable
Constructors
public PolInterp (double[] x, double[] y)
Constructs a new polynomial interpolating through the given points (x[0], y[0]), ...,
(x[n], y[n]). This constructs a polynomial of degree n from n+1 points.

Methods
public static double[] getCoefficients (double[] x, double[] y)
Computes and returns the coefficients the polynomial interpolating through the given points
(x[0], y[0]), ..., (x[n], y[n]). This polynomial has degree n and there are n+1 coefficients.

public double[] getX()

Returns the x coordinates of the interpolated points.

public double[] getY()

Returns the y coordinates of the interpolated points.

public static String toString (double[] x, double[] y)

Makes a string representation of a set of points.

public String toString()

Calls toString (double[], double[]) with the associated points.

September 29, 2015

LeastSquares
This class implements different linear regression models, using the least squares method
to estimate the regression coefficients. Given input data xij and response yi , one want to
find the coefficients j that minimize the residuals of the form (using matrix notation)
r = min kY Xk2 ,

where the L2 norm is used. Particular cases are

r = min

y i 0

k
X

!2
j xij

j=1

for k regressor variables xj . The well-known case of the single variable x is

X
r = min
(yi xi )2 .
,

Sometimes, one wants to use a basis of general functions j (t) with a minimization of
the form
!2
k
X
X
r = min
yi
j j (ti ) .

j=1

For example, we could have j (t) = ej t or some other functions. In that case, one has to
choose the points ti at which to compute the basis functions, and use a method below with
xij = j (ti ).
package umontreal.iro.lecuyer.functionfit;
public class LeastSquares
Methods
public static double[] calcCoefficients (double[] X, double[] Y)
Computes the regression coefficients using the least squares method. This is a simple linear
regression with 2 regression coefficients, and . The model is
y = + x.
Given the n data points (Xi , Yi ), i = 0, 1, . . . , (n 1), the method computes and returns the
array [, ].

public static double[] calcCoefficients (double[] X, double[] Y, int deg)

Computes the regression coefficients using the least squares method. This is a linear regression with a polynomial of degree deg = k and k + 1 regression coefficients j . The model
is
k
X
y = 0 +
j xj .
j=1

LeastSquares 4

September 29, 2015

Given the n data points (Xi , Yi ), i = 0, 1, . . . , (n 1), the method computes and returns the
array [0 , 1 , . . . , k ]. Restriction: n > k.

public static double[] calcCoefficients0 (double[][] X, double[] Y)

Computes the regression coefficients using the least squares method. This is a model for
multiple linear regression. There are k+1 regression coefficients j , and k regressors variables
xj . The model is
k
X
y = 0 +
j xj .
j=1

There are n data points Yi , Xij , i = 0, 1, . . . , (n 1), and each Xi is a k-dimensional

point. Given the response Y[i] and the regressor variables X[i][j], i = 0, 1, . . . , (n 1),
j = 0, 1, . . . , (k 1), the method computes and returns the array [0 , 1 , . . . , k ]. Restriction:
n > k + 1.

public static double[] calcCoefficients (double[][] X, double[] Y)

Computes the regression coefficients using the least squares method. This is a model for
multiple linear regression. There are k regression coefficients j , j = 0, 1, . . . , (k 1) and k
regressors variables xj . The model is
y=

k1
X

j xj .

j=0

There are n data points Yi , Xij , i = 0, 1, . . . , (n 1), and each Xi is a k-dimensional point.
Given the response Y[i] and the regressor variables X[i][j], i = 0, 1, . . . , (n 1), j =
0, 1, . . . , (k 1), the method computes and returns the array [0 , 1 , . . . , k1 ]. Restriction:
n > k.

September 29, 2015

BSpline
Represents a B-spline with control points at (Xi , Yi ). Let Pi = (Xi , Yi ), for i = 0, . . . , n
1, be a control point and let tj , for j = 0, . . . , m 1 be a knot. A B-spline [1] of degree
p = m n 1 is a parametric curve defined as
P(t) =

n1
X

Ni,p (t)Pi , for tp t tmp1 .

i=0

Here,
ti+p+1 t
t ti
Ni,p1 (t) +
Ni+1,p1 (t)
ti+p ti
ti+p+1 ti+1

1 for ti t ti+1 ,
Ni,0 (t) =
0 elsewhere.

Ni,p (t) =

This class provides methods to evaluate P(t) = (X(t), Y (t)) at any value of t, for a Bspline of any degree p 1. Note that the evaluate method of this class can be slow, since it
uses a root finder to determine the value of t for which X(t ) = x before it computes Y (t ).
package umontreal.iro.lecuyer.functionfit;
public class BSpline implements MathFunction
Constructors
public BSpline (final double[] x, final double[] y, final int degree)
Constructs a new uniform B-spline of degree degree with control points at (x[i], y[i]).
The knots of the resulting B-spline are set uniformly from x[0] to x[n-1].

public BSpline (final double[] x, final double[] y, final double[] knots)

Constructs a new uniform B-spline with control points at (x[i], y[i]), and knot vector
given by the array knots.

Methods
public double[] getX()
Returns the Xi coordinates for this spline.

public double[] getY()

Returns the Yi coordinates for this spline.

public double getMaxKnot()

Returns the knot maximal value.

BSpline 6

September 29, 2015

public double getMinKnot()

Returns the knot minimal value.

public double[] getKnots()

Returns an array containing the knot vector (t0 , tm1 ).

public static BSpline createInterpBSpline (double[] x, double[] y,

int degree)
Returns a B-spline curve of degree degree interpolating the (xi , yi ) points [1]. This method
uses the uniformly spaced method for interpolating points with a B-spline curve, and a
uniformed clamped knot vector, as described in http://www.cs.mtu.edu/~shene/COURSES/
cs3621/NOTES/.

public static BSpline createApproxBSpline (double[] x, double[] y,

int degree, int h)
Returns a B-spline curve of degree degree smoothing (xi , yi ), for i = 0, . . . , n points. The
precision depends on the parameter h: 1 degree h < n, which represents the number
of control points used by the new B-spline curve, minimizing the quadratic error
L=


n 
X
Yi Si (Xi ) 2
i=0

Wi

This method uses the uniformly spaced method for interpolating points with a B-spline curve
and a uniformed clamped knot vector, as described in http://www.cs.mtu.edu/~shene/
COURSES/cs3621/NOTES/.

public BSpline derivativeBSpline()

Returns the derivative B-spline object of the current variable. Using this function and the
returned object, instead of the derivative method, is strongly recommended if one wants
to compute many derivative values.

public BSpline derivativeBSpline (int i)

Returns the ith derivative B-spline object of the current variable; i must be less than the
degree of the original B-spline. Using this function and the returned object, instead of the
derivative method, is strongly recommended if one wants to compute many derivative
values.

September 29, 2015

SmoothingCubicSpline
Represents a cubic spline with nodes at (xi , yi ) computed with the smoothing cubic spline
algorithm of Schoenberg [1, 2]. A smoothing cubic spline is made of n + 1 cubic polynomials.
The ith polynomial of such a spline, for i = 1, . . . , n1, is defined as Si (x) while the complete
spline is defined as
S(x) = Si (x),
for x [xi1 , xi ].
For x < x0 and x > xn1 , the spline is not precisely defined, but this class performs extrapolation by using S0 and Sn linear polynomials. The algorithm which calculates the smoothing
spline is a generalization of the algorithm for an interpolating spline. Si is linked to Si+1 at
xi+1 and keeps continuity properties for first and second derivatives at this point, therefore
00
0
(xi+1 ).
(xi+1 ) and Si00 (xi+1 ) = Si+1
Si (xi+1 ) = Si+1 (xi+1 ), Si0 (xi+1 ) = Si+1
The spline is computed with a smoothing parameter [0, 1] which represents its accuracy with respect to the initial (xi , yi ) nodes. The smoothing spline minimizes
L=

n1
X

xn1

wi (yi Si (xi )) + (1 )

i=0

(S 00 (x)) dx

x0

In fact, by setting = 1, we obtain the interpolating spline; and we obtain a linear function
by setting = 0. The weights wi > 0, which default to 1, can be used to change the
contribution of each point in the error term. A large value wi will give a large weight to the
ith point, so the spline will pass closer to it. Here is a small example that uses smoothing
splines:
int n;
double[] X = new double[n];
double[] Y = new double[n];
// here, fill arrays X and Y with n data points (x_i, y_i)
// The points must be sorted with respect to x_i.
double rho = 0.1;
SmoothingCubicSpline fit = new SmoothingCubicSpline(X, Y, rho);
int m = 40;
double[] Xp = new double[m+1];
double[] Yp = new double[m+1];
double h = (X[n-1] - X[0]) / m;
for (int i = 0; i <= m; i++) {
double z = X[0] + i * h;
Xp[i] = z;
Yp[i] = fit.evaluate(z);
}

// Xp, Yp are spline points

// step

// evaluate spline at z

September 29, 2015

SmoothingCubicSpline 8

package umontreal.iro.lecuyer.functionfit;
import umontreal.iro.lecuyer.functions.*;
import umontreal.iro.lecuyer.functions.Polynomial;
public class SmoothingCubicSpline implements MathFunction,
MathFunctionWithFirstDerivative, MathFunctionWithDerivative,
MathFunctionWithIntegral
Constructors
public SmoothingCubicSpline (double[] x, double[] y, double[] w,
double rho)
Constructs a spline with nodes at (xi , yi ), with weights wi and smoothing factor = rho.
The xi must be sorted in increasing order.

public SmoothingCubicSpline (double[] x, double[] y, double rho)

Constructs a spline with nodes at (xi , yi ), with weights = 1 and smoothing factor = rho.
The xi must be sorted in increasing order.

Methods
public double evaluate (double z)
Evaluates and returns the value of the spline at z.

public double integral (double a, double b)

Evaluates and returns the value of the integral of the spline from a to b.

public double derivative (double z)

Evaluates and returns the value of the first derivative of the spline at z.

public double derivative (double z, int n)

Evaluates and returns the value of the n-th derivative of the spline at z.

public double[] getX()

Returns the xi coordinates for this spline.

public double[] getY()

Returns the yi coordinates for this spline.

public double[] getWeights()

Returns the weights of the points.

public double getRho()

Returns the smoothing factor used to construct the spline.

public Polynomial[] getSplinePolynomials()

Returns a table containing all fitting polynomials.

September 29, 2015

SmoothingCubicSpline 9

public int getFitPolynomialIndex (double x)

Returns the index of P , the Polynomial instance used to evaluate x, in an ArrayList table
instance returned by getSplinePolynomials(). This index k gives also the interval in table
X which contains the value x (i.e. such that xk < x xk+1 ).

September 29, 2015

REFERENCES 10

References
[1] C. de Boor. A Practical Guide to Splines. Number 27 in Applied Mathematical Sciences
Series. Springer-Verlag, New York, 1978.
[2] D. S. G. Pollock. Smoothing with cubic splines. Technical report, University of London,
Queen Mary and Westfield College, London, 1993.