pweglik / badania-operacyjne Goto Github PK

Create the graph of the real city. Remember about representation we agreed on (junctions and stops are vertices, junctions with value 0 and stops with preprocessed value, also stops have degree <= 2).

Or at least make current generated graph meet the aforementioned conditions.

implement some variations of the score function

when #17 and maybe others are done do grid search of the hiperparameters
Hiperparameters to optimize:

• survival function
• mutations probabilities
• sanitizers

Do that for many fitness functions, but do not compare results numerically.

In Python 3.9+ you don't need to import List,Tuple etc. but you can use list, tuple etc. as type annotations.

add metrics to allow monitoring:

• number of mutaions (per type)
• number of crosses (per type)
• more...

Similary to pandas and scikit learn we may consider creating _latest_gen variable inside of SImulationEngine so we can access that generaion after simulation and play with it, dispaly lines separately etc.

Determine type of best_paths that is used everywhere and type its usages. Ideally use type aliases to not repeat weird nested type in N places.

When no edges are present in the graph this line throws:

ValueError: max() arg is an empty sequence

Genotype was 1 line with 1 stop.

This needs to be described and fixed

This should cause more % of the graph to by covered by lines

This somehow prints different results:

# calculating fitness for all organisms
population_with_fitness: List[Tuple[Genotype, float]] = [
    (organism, self.fitness_function(organism, self.G))
    for organism in population
]

print(population_with_fitness[0][1])

population_with_fitness: List[Tuple[Genotype, float]] = [
    (organism, self.fitness_function(organism, self.G))
    for organism in population
]

print(population_with_fitness[0][1])

This is related:

G.graph["points"][np.where(lines_stopping_count == 0)] = 0
lines_stopping_count[np.where(lines_stopping_count == 0)] = 1

Merytoryczne

• Funkcja K(Si) opisać co robi
• Aplikacja
• Jak wygenerowalsmy odległości ścieżek (Floyd Warshal)
• Jak z przystanków są krawedzie wygenerowane

Typing

• usunąć puste pola
• Poprawić ortograficzne błędy
• Źródła i podsumowanie

implement some variations of survival function
ideas:

• add some randomness so bad specimens can get lucky

naprawić osmnx config option
tak żeby działały notebooki
dla random mapy
dla mapy krk
grid search