Multibrot Set

2017-04-22, post № 168

Java, mathematics, programming, #animated gif, #animation, #Cartesian, #complex, #complex arithmetic, #fractal, #generalization, #gif, #Mandelbrot set, #multi-threading, #polar, #reals, #threading

The Mandelbrot Set is typically defined as the set of all numbers c\in\mathbb{C} for which — with z_0=0, z_{n+1}=f_c(z_n) and f_c(z)=z^2+c — the limit \lim\limits_{n\to\infty}z_n converges. Visualizations of this standard Mandelbrot Set can be seen in three of my posts (Mandelbrot Set, Mandelbrot Set Miscalculations and Mandelbrot Set II).


However, one can extend the fractal’s definition beyond only having the exponent 𝟤 in the function to be f_c(z)=z^\text{exp}+c with \text{exp}\in\mathbb{R} [1]. The third post I mentioned actually has some generalization as it allows for \text{exp}\in\{2,3,4,5\}, although the approach used cannot be extended to real or even rational numbers.


The method I used in the aforementioned post consists of manually expanding (a+b\cdot i)^n for each 𝑛. The polynomial (a+b\cdot i)^3, for example, would be expanded to (a^3-3\cdot a\cdot b^2)+(3\cdot a^2\cdot b-b^3)\cdot i.
This method is not only tedious, error-prone and has to be done for every exponent (of which there are many), it also only works for whole-number exponents. To visualize real Multibrots, I had to come up with an algorithm for complex number exponentiation.


Luckily enough, there are two main ways to represent a complex number, Cartesian form z=a+b\cdot i and polar form z=k\cdot e^{\alpha\cdot i}. Converting from Cartesian to polar form is simply done by finding the number’s vector’s magnitude k=\sqrt{a^2+b^2} and its angle to the 𝑥-axis \alpha=\mbox{atan2}(\frac{a}{b}). (The function \mbox{atan2} is used in favor of \arctan to avoid having to divide by zero. View this Wikipedia article for more on the function and its definition.)
Once having converted the number to polar form, exponentiation becomes easy, as

z^\text{exp}=(k\cdot e^{\alpha\cdot i})^\text{exp}=k^\text{exp}\cdot e^{\alpha\cdot\text{exp} \cdot i}.

With the exponentiated z^\text{exp} in polar form, it can be converted back in Cartesian form with

z^\text{exp}=k^\text{exp}\cdot (\cos{(\alpha\cdot\text{exp})}+\sin{(\alpha\cdot\text{exp})}\cdot i\big).

Using this method, converting the complex number to perform exponentiation, I wrote a Java program which visualizes the Multibrot for a given range of exponents and a number of frames.
Additionally, I added a new strategy for coloring the Multibrot Set, which consists of choosing a few anchor colors and then linearly interpolating the red, green and blue values. The resulting images have a reproducible (in contrast to randomly choosing colors) and more interesting (in contrast to only varying brightness) look.


The family of Multibrot Sets can also be visualized as an animation, showing the fractal with an increasing exponent. The animated gif shown below was created using ImageMagick’s convert -delay <ms> *.png multibrot.gif command to stitch together the various .png files the Java application creates. To speed up the rendering, a separate thread is created for each frame, often resulting in 𝟣𝟢𝟢% CPU-usage. (Be aware of this should you render your own Multibrot Sets!)


To use the program on your own, either copy the source code listed below or download the .java file. The sections to change parameters or the color palette are clearly highlighted using block comments (simply search for /*).
To compile and execute the Java application, run (on Linux or MacOS) the command javac multibrot.java; java -Xmx4096m multibrot in the source code’s directory (-Xmx4096m tag optional, though for many frames at high quality it may be necessary as it allows Java [2] to use more memory).
If you are a sole Windows user, I recommend installing the Windows 10 Bash Shell.

Source code: multibrot.java


2017-04-16, post № 167

art, ascii, haiku, poetry, #ascii egg, #celebration, #easter egg, #egg

Winds swirl through the air,
Water sloshes at the shore;
A peaceful island.


T-3PO — Tic-Tac-Toe Played Optimally

2017-04-08, post № 166

games, HTML, JavaScript, mathematics, programming, Python, #computer player, #game ai, #perfect, #perfect play, #three in a row, #three to win, #three

Tic-Tac-Toe, noughts and crosses, Xs and Os, three in a row or whatever you want to call it may be the simplest perfect information game that is enjoyable by humans. Two players set their pieces (X or O) on an 𝟥 ⨉ 𝟥 grid, alternating their turns. The first player to get three of their pieces in a line, wins. If no player succeeds to get a line, the game ends in a draw.

Tic-Tac-Toe’s simplicity may become clear, if you consider that skilled players — people who have played a few rounds — can reliably achieve a draw, thereby playing perfectly. Two perfect players playing Tic-Tac-Toe will — whoever starts — always tie, so one may call the game virtually pointless, due to there practically never being a winner.
Because of its simple rules and short maximal number of turns (nine) it is also a game that can be solved by a computer using brute-force and trees.

The first Tic-Tac-Toe-playing program I wrote is a Python shell script. It lets you, the human player, make the first move and then calculates the best possible move for itself, leading to it never loosing. On its way it has a little chat whilst pretending to think about its next move. The Python source code can be seen below or downloaded here.

The second Tic-Tac-Toe-playing program I wrote uses the exact same method of optimizing its play, though it lets you decide who should begin and is entirely written in JavaScript. You can play against it by following this link.

Both programs look at the entire space of possible games based on the current board’s status, assumes you want to win and randomly picks between the moves that either lead to a win for the computer or to a draw. I did not include random mistakes to give the human player any chance of winning against the computer. Other Tic-Tac-Toe-playing computers, such as Google’s (just google the game [1]), have this functionality.

Source code: t-3po.py

Second Anniversary

2017-03-28, post № 165

art, #2 years, #celebration, #collage, #j, #J-Blog, #jblog, #two years

J-Blog celebrates its second anniversary!
Exactly two years ago, on the twenty-eighth of March 2015, the very first post on this blog appeared, appropriately named “Hello World”. Since then, including this one, 𝟣𝟨𝟦 other posts have been posted. Here a few of the image highlights from both years, with their corresponding post.

       JW&Mbw\1] ZB@b    
       ?-        #BB%c   
Ckc                k&z   
b@#Z?            [a8L    
X%$@*Q[?      ?[q%&}     

Bifurcation Diagram

2017-03-25, post № 164

mathematics, PIL, programming, Python, #alpha, #chaos, #chaos theory, #delta, #Feigenbaum, #fractal, #iterations, #Mandelbrot set, #modelling, #population

Generating the famous fractal, which can be used to model populations with various cycles, generate pseudo-random numbers and determine one of nature’s fundamental constants, the Feigenbaum constant 𝛿.
The fractal nature comes from iteratively applying a simple function, x\mapsto\lambda\cdot x\cdot (1-x) with 0\leq\lambda\leq 4, and looking at its poles.
The resulting image looks mundane at first, when looking at 0\leq\lambda\leq 3, though the last quarter section is where the interesting things are happening (hence the image below only shows the diagram for 2\leq\lambda\leq 4).
From 𝜆 = 𝟥 on, the diagram bifurcates, always doubling its number of poles, until it enters the beautiful realm of chaos and fractals.


For more on bifurcation, fractals and 𝛿, I refer to this Wikipedia entry and WolframMathworld.

Source code: bifurcation-diagram.py


2017-03-14, post № 163

mathematics, programming, Python, #approximation, #dimensions, #four, #four dimensions, #generator, #higher dimensions, #hyperspheres

Every year on March the 14th, for one day the world gets irrationally excited about the famous constant 𝜋. As is tradition, you try to calculate 𝜋 in unusual ways, demonstrating the constant’s ubiquity as it crops up in the most unexpected circumstances.

  u$$mnn:       Y$$$i        .@$$$$$,             
 $$n           )$$$*         W$$$$$m              
-n[            $$$$.        ]$$$$$$               
              h$$$w         $$$$$$Y               
             [$$$$         X$$$$$$                
            "$$$$n        '$$$$$${                
           .$$$$$         8$$$$$$                 
           *$$$$}        :$$$$$$+                 
          #$$$$u         $$$$$$%                  
        t$$$$$$         ;$$$$$$`                  
       u$$$$$$!         $$$$$$W                   
      Y$$$$$$M         .$$$$$$,                   
    f$$$$$$$$.         Z$$$$$Z          nn        
  `w$$$$$$$$|          $$$$$$(         v$z        
 n$$$$$$$$$W           $$$$$$$1      'X$8         
Y$$$$$$$$$$            *$$$$$$$8nnnn$$$p          
$$$$$$$$$@.             W$$$$$$$$$$$$$n           
_$$$$$$${                x$$$$$$$$$0>             
   -n{.                     !|nt_.                

A fairly well-known way to approximate 𝜋 is to randomly choose points in a square (often thought of as throwing darts at a square piece of cardboard), determine their distance to a circle’s center and do a division, as I did in my 𝜋 Generator post.

However, 𝜋 does not only appear in the formula for a circle’s area, A=\pi\,r^2, yet [1] also in the formula for a sphere’s volume, V=\frac{4}{3}\,\pi\,r^3, and for all the infinite hyperspheres above dimension three (view this Wikipedia article for more about volumes of higher-dimensional spheres).

In particular, the formula for the hypervolume of a hypersphere in four dimensions is defined as being V=\frac{\pi^2}{2} \cdot r^4. Using this formula, my Python script randomly chooses four-dimensional points (each in the interval \left[0,1\right)), calculates their distance to the point \left(0.5,0.5,0.5,0.5\right) and determines if they are in the hypersphere around that point with radius 𝟢.𝟧.
By dividing the number of random points which lie in the hypersphere by the number of iterations used (10^6 in the example below), the script approximates the hypersphere’s hypervolume. By then rearranging the equation V=\frac{\pi^2}{2}\cdot r^4 with 𝑟 = 𝟢.𝟧 to \pi=\sqrt{V\cdot 32}, the desired constant can be approximated.

$ python pi.py
Source code: pi-day-mmxvii.py


2017-03-11, post № 162

mathematics, programming, Python, #2016, #decimal reverse, #difference, #integer, #OEIS, #On-Line Encyclopedia of Integer Sequences, #palindrome, #reverse, #sequence, #square

The On-Line Encyclopedia of Integer Sequences (also known by its acronym, OEIS) is a database hosting hundreds of thousands of — as the name implies — integer sequences. Yet, despite the massive number of entries, I contributed a new integer sequence, A278328.

A278328 describes numbers whose absolute difference to their decimal reverse are square. An example would be 𝟣𝟤 or 𝟤𝟣 (both are the decimal reverse to each other), since \left|12-21\right|=9 and 9=3^2.

Not a whole lot is known about the sequence [1], partly due to its definition only resulting in the sequence when using the decimal system, though it is known that there are infinitely many numbers with said property. Since there are infinitely many palindromes (numbers whose reverse is the number itself), \left|n-n\right|=0 and 0=0^2.

Due to there — to my knowledge — not being a direct formula for those numbers, I wrote a Python script to generate them. On the sequence’s page, I posted a program which endlessly spews them out, though I later wrote a Python two-liner, which only calculates those members of the sequence in the range from 𝟢 to 𝟫𝟪 (shown below entered in a Python shell).

>>> import math
>>> filter(lambda n:math.sqrt(abs(n-int(str(n)[::-1])))%1 == 0, range(99))
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 15, 21, 22, 23, 26, 32, 33, 34, 37, 40, 43, 44, 45, 48, 51, 54, 55, 56, 59, 62, 65, 66, 67, 73, 76, 77, 78, 84, 87, 88, 89, 90, 95, 98]

Maze Solving

2017-02-27, post № 161

programming, Python PIL, #amazing, #labyrinth, #Minotaur, #solver

Mazes have been a subject of human interest for thousands of years. The Greeks used them to trap a bull-man hybrid, the French built them to show how they could impose order on nature, and even nowadays people enjoy wandering around corn mazes.
The algorithmic art of using computers to solve mazes — and even to find the shortest path through a maze —, however, has only emerged in the last couple of decades.

I was inspired by a recent Computerphile video in which Michael Pound talks about implementing different path finding algorithms for use in maze solving. And as he used Python — one of my favourite languages out there [1] —, I thought I could give it a try and came up with this maze solver.


The mazes given to the solver (through a .png file) have to have a specific form. The maze needs to have a border all around (painted black) with two holes at the top and bottom, marking the maze’s start and exit (all path pixels are white).
Then the solver — using PIL — reads in the maze file, determines start and exit and starts at the maze’s start, labelling each maze path according to its shortest distance to the start. After it has found the exit, it stops looking at the maze and traces its origins back from the exit, marking the path it goes along as the maze’s optimal solution (highlighted in red).
The different hues of blue indicate the tile’s distance to the start, the white tiles are tiles the solver did not even look at.The different shadings also reveal information about the maze. Mazes with only one solution tend to have sharp changes as there are parts of the maze separated by only one wall, yet separated by a huge walk distance through the maze. The one maze with multiple solutions (see upper right image below) — in contrast — has only gradual changes in hue.

To solve a 𝟦 megapixel maze, the solver takes around 𝟥 seconds, for a 𝟣𝟨 megapixel maze around 𝟣𝟦 seconds and for a 𝟤𝟤𝟧 megapixel maze around 𝟩 minutes and 𝟤𝟤 seconds.
Performance was measured on an Intel Core i7 (𝟦.𝟢𝟢 GHz).

All mazes shown were downloaded from Michael Pound’s mazesolving GitHub repository, which were mostly generated using Daedalus.

The solver’s source code is listed below, though you can also download the .py file.

Source code: maze-solving.py
Jonathan Frech's blog; built 2024/07/06 12:42:51 CEST