Find a good alphabet with genetic algorithm
- choose the current alphabet (optimized by letters frequency) and calculate its value*;
- for all possible 2-letter shifting permutations (2625 = 650), get the 10% best alphabets based on their value;
- repeat step 2.
- wait.
*processing the value of an alphabet:
- for each word in the top-1000 most used word, based on the english words frequency, calculate its value:
- sum of each edge length (1000, 1414 or 2236);
- malus -10000 for each component of the graph;
- malus -1000 for each intersection;
- malus -100 for each turns;
- multiply this value by the number of iteration of this word in the frequency list (divided by 1000 to avoid large numbers);
- return the sum of all the words values, divided by a fixed value to avoid large numbers.
What is better?
- points number: minimum, maximum, distributed
- position : top, bottom, center, distributed
Edited by Nathanaël