Assignment 4 Interpolation

CS 370 Numerical Computation Assignment 4
Interpolation
What you need to get
• https://www.overleaf.com/read/vztpkwrpmtcn: (optional) Overleaf template for Q1 and Q2
• YOU_a4q3.ipynb
• YOU_a4q4.ipynb
What to do
1. [5 marks] Consider the following alternative representation for a cubic spline, S(x):
S(x) =



20 + 25x + 18x
2 + a1x
3 −3 ≤ x < −1
26 + a2x + a3x
2 + a4x
3 −1 ≤ x < 0
26 + 19x + a5x
2 + x
3 0 ≤ x ≤ 3.
Is it possible to find values for {a1, a2, a3, a4, a5} so that S(x) is a natural cubic spline? Justify your
answer.
2. [5 marks] The figure shows four samples of a function f on a 2D domain. However, you would like
to interpolate it at (x, y) to get a value fx,y, where
0 < x < 1 and 0 < y < 1. There are three strategies
for doing this:
(a) Interpolate horizontally between f0,1 and
f1,1 to get a, interpolate horizontally between f0,0 and f1,0 to get c, and then interpolate vertically between those two values to
get the value fx,y at (x, y).
(b) Interpolate vertically between f0,0 and f0,1
to get d, interpolate vertically between f1,0
and f1,1 to get b, and then interpolate horizontally between those two values to get the
value fx.y at (x, y).
a
b
c
d
f0,0 = 12 f1,0 = 15
f0,1 = 8 f1,1 = 10
fx,y
(c) Take a weighted sum of the f-values, where the weight for each f-value is given by the area
of the diagonally opposite rectangle. For example, the weight for f0,1 is (1 − x)y.
Prove that all three methods give the same value fx,y. Show your work.
3. [7 marks] Complete the Python function MySpline that reads in a set of x and y values (each
as arrays or lists), and outputs a piecewise-cubic function with natural boundary conditions. The
function prototype is
cs = MySpline(x, y)
The returned function can be called using cs(2.4) or cs([2.4, 3.7]), for example. You will
have to find the parameter values for the as, bs, and cs so that the cubic spline can be evaluated
c Jeff Orchard, Reinhold Burger 2019 v1.0 Page 1
CS 370 Numerical Computation Assignment 4
using
pi(x) = ai
(xi+1 − x)
3
6hi
+ ai+1
(x − xi)
3
6hi
+ bk (xi+1 − x) + ci (x − xi) ,
where hi = xi+1 − xi and i = 1, 2, . . . , n − 1. Note that all Python indexing starts at 0, so all the
indices can be decremented by 1. However, some care is needed to handle a1, . . . , an. See the
documentation for MySpline for some guidance on this issue.
The function MySpline is already set up to return a cubic spline function. But it’s not a very
interesting one, and it does not pass through the points (xi
, yi).
The notebook has a small sample set of points to interpolate. Feel free to choose your own set of
points. Create a figure that plots the interpolation points overtop of the smooth cubic spline.
4. [8 marks] Create a parametric curve representation of your nickname written in your handwriting.
This representation should be based on parametric spline interpolation. Your nickname must have
at least two curved segments. Figure 1 below shows an example using 5 coarsly-sampled segments.
Figure 1: The nickname “T-Bone” written with 5 parametric cubic spline segments
Steps:
(a) Draw your nickname and select interpolation points. You can do this on graph paper by hand,
or you can use ginput. Make sure the interpolation points are not too close together. Notice
that the loops of the “B” in Figure 1 are made from only 6 points each. Your curved segments
should be sampled sparingly.
(b) Add code to the notebook to initialize and store the interpolation points for each segment.
For example, you could store the coordinates of the interpolation points for the first segment
in the arrays x1 and y1, and you would have one x-y pair of arrays for each segment. Your
data arrays should be hardcoded into your script.
(c) Complete the function ParametricSpline; it takes the interpolation points of one segment
as input, and outputs a pair of cubic spline functions (one for x(t) and one for y(t)), along
with an array, t, that holds the parameter value at the interpolation points. See the function’s
documentation for more details. The parameter t must reflect the pseudo-arclength of the
curve. You can use scipy.interpolate.make_interp_spline (and return the spline
functions that it creates). Alternatively, if you feel confident in your MySpline function, you
can use that instead.
(d) Add lines to the notebook to call ParametricSpline for each segment in your nickname.
(e) Add more lines to to the notebook to plot your nickname. The plot should graph the parametric spline segments using a refined selection of parameter values (eg. 1000 points per
c Jeff Orchard, Reinhold Burger 2019 v1.0 Page 2
CS 370 Numerical Computation Assignment 4
segment). The plot should also display the original interpolation points, and the x and y axes
should be rendered using the same scale (“plt.axis(’equal’)”). Figure 1 is an example
of what your output should look like.
What to submit
You must submit a series of PDF documents to Crowdmark. For each coding question, export your
jupyter notebook to PDF. When a proof or manual calculation is requested, you can typeset your solution
(eg. using LATEX or Word), or write your solution by hand. In either case, your solution should also be
submitted as a PDF. If you submit photos of handwritten work, it is your responsibility to ensure that
they are legible.
c Jeff Orchard, Reinhold Burger 2019 v1.0 Page 3