22.01.2015 Views

Refined Buneman Trees

Refined Buneman Trees

Refined Buneman Trees

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

100000<br />

90000<br />

Performance of the refined <strong>Buneman</strong> tree algorithm<br />

best fit<br />

80000<br />

70000<br />

Running time (ms.)<br />

60000<br />

50000<br />

40000<br />

30000<br />

20000<br />

10000<br />

0<br />

4 8 16 32 64 128<br />

Input size<br />

Figure 15.1: The running time of the refined <strong>Buneman</strong> algorithm. Notice the<br />

artifact at power of two intervals. This stems from the quad tree data structure<br />

described in chapter 8<br />

• Input stop size 174 1<br />

• Number of repetitions 100<br />

To analyse the asymptotic running time performance of the algorithm, the<br />

test data was analysed using the nonlinear least-squares (NLLS) Marquardt-<br />

Levenberg algorithm via the gnuplot ([gnu99]) fit command.<br />

The expectation was that the performance of the algorithm would be O(n 3 ),<br />

and therefore the fit-function was chosen to be f(x) = ax b . The fit of the<br />

function f with the whole dataset (input sizes 4 – 174) returned the following<br />

values: a =0.0130805 and b =3.02047. A plot of the data along with the<br />

function f is shown in Figure 15.1. The figure shows a plot of the running time<br />

in milliseconds against the input size in number of species.<br />

Figure 15.1 shows an artifact from the implementation, namely a jump in<br />

running time performance consistent with intervals matching powers of 2. This<br />

artifact stems from the dimensioning of the quad tree data structure described in<br />

1 Actually the stop size was 200, but after 3 consecutive days of running the test it was<br />

halted due to impatience.<br />

82

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!