Fractals generated by randomly initialized neural networks.
To use run:
python fractal.py
You'll be prompted with an input asking for a random seed, press enter if you just want to use a random one. A random fractal will be shown to you, probably at a low resolution. The reason for the low resolution is because fractal generation is compute intensive, as you need to compute the trajectory of every point on the complex plane; the fact that a neural network is involved greatly increases the compute.
Click on a region to zoom in on it, be careful, it may take some time. To refine the image (increase resolution), type
refine #resolution_scale
where #resolution_scale is the factor we increase the resolution by (2 would bring a 400 x 400 resolution to a 800 x 800 resolution). This command may take a lot of time to compute, because we need to completely re-graph the function. Working on a faster way of doing this.
To perform more iterations (adds more detail to the fractal), type
iterate #num_iterations
where #num_iterations is the number of extra iterations we perform.