Parallelization of the Box Counting method - Guy Tel-Zur

Download Report

Transcript Parallelization of the Box Counting method - Guy Tel-Zur

Ravid Rodney 066514837
Or Maltabashi 308551332
Outlines
• What is Fractal?
• History
• Fractal dimension
• Box Counting Method
• Fractal dimension Calculations:
Version 1
Version 2
• Parallelization of the Box Counting method
What is Fractal?
• The word fractal comes from the Latin “Fract-” which means broken.
• Fractal is a geometric pattern that is repeated at ever smallest
scales to produce irregular shapes and surface that cannot be
represented by classical geometry.
• Fractals are used especially in computer modeling of irregular
patterns and structure in nature.
History
• The mathematics behind fractals began to take shape in the
17𝑡ℎ century when Gottfried Leibniz pondered recursive self-similarity.
• In 1872 Katl Weierstrass presented the first definition of a function with
a graph that would today be considered fractal.
• In 1915, Sierpiński constructed his famous triangle fractal.
• In March 1918, Felix Hausdorff expanded the definition of “fractal
dimension“.
• In 1975, Mandelbort coined the word “fractal” and illustrated his
mathematical definition with striking computerconstructed visualizations.
Fractal Dimension
• There are two definitions of fractal dimension:
● Fractal dimension is a ratio providing a statistical index of
complexity comparing how detail in a pattern changes with the
scale at which it is measured.
● Fractal dimension is a measure of the space-filling capacity of a
pattern that tells how a fractal scales differently than the space
it is embedded in.
• A fractal dimension does not have to be
an integer.
Box Counting Method
• The method of Box Counting is a way of gathering data from a
complex pattern or image by breaking it up into smaller and smaller
pieces, and analyzing each piece separately.
• The essence of the process has been compared to zooming in or
out using optical or computer based
methods to examine how observations
of detail change with scale.
• This method is embarrassingly parallel!
Box Counting Method – Fractal
Dimension Calculation Ver. 1
• In order to calculate fractal’s dimension you begin by covering the
fractal image area with different grid sizes. Then count the number
of grid blocks containing part of the fractal in them.
◦
◦ S=1/12
S=1/3
Box Counting Method – Fractal
Dimension Calculation Ver. 1
• After counting a sufficient amount of grid sizes we calculate the
Fractal dimension using the formula:
• In the previous example, Where n() is the number of grid blocks
containing a part of the fractal and 1/S is the grid scale.
• 𝐷𝑔𝑟𝑖𝑑1𝑥 1 =
3 12
𝑙𝑜𝑔105−𝑙𝑜𝑔18
𝑙𝑜𝑔12−𝑙𝑜𝑔3
≈ 1.27 or 𝐷𝑔𝑟𝑖𝑑1𝑥1 =
3 6
𝑙𝑜𝑔48−𝑙𝑜𝑔18
𝑙𝑜𝑔6−𝑙𝑜𝑔3
≈ 1.19
Which if we cont. to more measurements will all average to 1.26 which
is the calculated dimension of the Koch Fractal.
Box Counting Method – Fractal
Dimension Calculation Ver. 2
• This method is slightly different from Ver. 1 in that it only calculates
the smallest grid size, then uses the same formula with the biggest
grid available (S=1 ,N=1) which leads us to this final formula:
𝐷𝑔𝑟𝑖𝑑_𝑠𝑚𝑎𝑙𝑙𝑒𝑠𝑡
𝑙𝑜𝑔𝑁 − 𝑙𝑜𝑔1
𝑙𝑜𝑔𝑁
=
=
1
1
log
− 𝑙𝑜𝑔1 log
𝑠
𝑠
Parallelization of the Box Counting
method
• For each version there is a different parallelization method from our
curriculum we can implement:
• For version 1, we’d use a Condor Approach:
• Then another process will take all the measurements, put them in
the formula and average the results.
Parallelization of the Box Counting
method
• For Version 2, we will parallelize using MPI or OpenMP :
• Serial Pseudo Code:
Parallelization of the Box Counting
method
• Parallel Pseudo Code:
Improvement of Parallelization of the
Box Counting method
• Yu-Chang and Kuo-Tai [1] tested an improved Parallel box counting
algorithm using version 2 with these results:
Improvement of Parallelization of the
Box Counting method
• Speedup:
Bibliography
• [1] A Parallel Differential Box-Counting Algorithm Applied to
Hyperspectral Image Classification : Yu-Chang Tzeng, Kuo-Tai Fan,
and Kun-Shan Chen
• [2] Using the Fractal Dimension to Cluster Datasets : Daniel
Barbar´a & Ping Chen
Questions?