This repository contains two strategies for printing Hello, world!:
- Co-called genetic algorithm
- Machine learning like approach
Bascially two approaches have a single input: randomly generated string
out of which each algorithm must derive target string, Hello, world!.
This algorithm described numerous times in various blogs and has different flavours. Logic behind this algorithm is simple: get a population and mutate it until cost function will output zero.
This algorithm can get complex with all the mimicry to evolution processes. In essense it mutates input string/strings unless in a way that cost function gets better and better results.
This approach relies on a linear transformation. Having a target and a random input string, of the same length as the target, algorithm find such a linear transformation (function) which converts the input string into the target string.
Linear transformation can be found numerically via change in data representation. String makes more sense in a shape of array of integers.
On this plot target represents a target string, Hello, world!, sample represents
a randomly generated string, subject to transform somehow into the target string.
Dashed line, weights, represents a vector. Using simple transformation function
f(x, w) = x + w
the sample string, x, will be transformed to target string via weight, w,
transformation.
This idea can be seen clearly on the plot where x = 2. Sample at x is slightly
above target at x, and weights at x is equally slightly below zero line.
Therefore, after transformation sample at x will be equal to target at x.