def compare_methods(g,
nbr_tests=default_tests,
p=default_p,
gamma=default_gamma):
d_multi_obj, v, t, time = algos.pl_part3_1(g, p, gamma)
d_pdm, v, t, time = algos.iteration_de_la_valeur(g, p, gamma)
# on calcul le cout moyen pour chacune des ces politiques, en pratique
mean_multi_obj = ot.mean_experienced_cost_p3(d_multi_obj, g, p, nbr_tests)
mean_pdm = ot.mean_experienced_cost_p3(d_pdm, g, p, nbr_tests)
return (mean_multi_obj, mean_pdm)
def affiche_res_iteration_valeur(self):
p = 0.6
gamma = 0.9
epsilon = 0.00001
d, val_etats, t, time = algos.iteration_de_la_valeur(
self.g, p, gamma, epsilon)
self.resultat.config(state=tk.NORMAL)
self.resultat.delete(1.0, tk.END)
self.resultat.insert(tk.INSERT, ot.from_action_to_dir(d, self.g))
self.resultat.insert(tk.INSERT, '\nnumber of iteration :' + str(t))
self.resultat.insert(tk.INSERT, '\ntime :' + str(time))
self.resultat.config(state=tk.DISABLED)
def impact_q(variation_q,
nbLignes=default_nbLignes,
nbCol=default_nbCol,
p=default_p,
gamma=default_gamma,
epsilon=default_epsilon):
couts = [0, 1, 2, 3, 4]
for q in variation_q:
random.seed(1)
np.random.seed(1)
print(q)
new_couts = [x**q for x in couts]
g = grille.Grille(nbLignes, nbCol, 2, 0.2, 0.2, 0.2, 0.2, 0.2,
new_couts, 1000).g
d, v, t, time = algos.iteration_de_la_valeur(g, p, gamma, epsilon)
print(ot.from_action_to_dir(d, g))
def resolution_nbr_cases(nbLignes=default_nbLignes,
nbCol=default_nbCol,
p=default_p,
gamma=default_gamma,
epsilon=default_epsilon):
size = nbLignes * nbCol
couts = [0, 1, size**1, size**2, size**3]
reward = 1000
grid = grille.Grille(nbLignes, nbCol, 2, 0.2, 0.2, 0.2, 0.2, 0.2, couts,
reward)
d, v, t, time = algos.iteration_de_la_valeur(grid.g, p, gamma, epsilon)
print(ot.from_action_to_dir(d, grid.g))
#affichage dans la GUI
grid.resultat.config(state=tk.NORMAL)
grid.resultat.delete(1.0, tk.END)
grid.resultat.insert(tk.INSERT, ot.from_action_to_dir(d, grid.g))
grid.resultat.insert(tk.INSERT, '\nnumber of iteration :' + str(t))
grid.resultat.insert(tk.INSERT, '\ntime :' + str(time))
grid.resultat.config(state=tk.DISABLED)
grid.Mafenetre.mainloop()
# resolution_nbr_cases()
def impact_size(sizes,
nbr_tests=default_tests,
p=default_p,
gamma=default_gamma,
epsilon=default_epsilon):
df = pd.DataFrame(columns=[
"dimensions",
"taille",
"iter_val",
"iter_pol",
"iter_pl",
"time_val",
"time_pol",
"time_pl",
])
for i, taille in enumerate(sizes):
print(taille)
somme1_it = 0
somme1_t = 0
somme2_it = 0
somme2_t = 0
somme3_it = 0
somme3_t = 0
for k in range(nbr_tests):
fenetre = grille.Grille(taille[0], taille[1], 2, 0.2, 0.2, 0.2,
0.2, 0.2, [0, 1, 2, 3, 4], 1000)
g = fenetre.g
# iteration de la valeur
d, v, t, time = algos.iteration_de_la_valeur(g, p, gamma, epsilon)
somme1_it += t
somme1_t += time
# iteration de la politique
d, v, t, time = algos.iteration_de_la_politique(
g, p, gamma, epsilon)
somme2_it += t
somme2_t += time
#pl
d, v, t, time = algos.pl(g, p, gamma)
somme3_it += t
somme3_t += time
# ajouter le pl
line = pd.Series({
'dimensions': taille,
'taille': taille[0] * taille[1],
'iter_val': somme1_it / nbr_tests,
'iter_pol': somme2_it / nbr_tests,
'iter_pl': somme3_it / nbr_tests,
'time_val': somme1_t / nbr_tests,
'time_pol': somme2_t / nbr_tests,
'time_pl': somme3_t / nbr_tests,
})
df = df.append(line, ignore_index=True)
# Affichage des résultats
print(df)
#stocke dans un fichier csv
df.to_csv("./impact_size.csv", ';')
plt.figure(1)
ax1 = plt.subplot(211)
plt.scatter(df.taille, df.iter_val, label="Iteration de la valeur")
plt.scatter(df.taille, df.iter_pol, label="Iteration de la politique")
plt.scatter(df.taille, df.iter_pl, label="Programmation mathématique")
plt.xlabel("Taille de l'instance")
plt.ylabel("Nombre d'itérations")
plt.setp(ax1.get_xticklabels(), visible=False)
plt.legend(bbox_to_anchor=(1, 1), loc=0, borderaxespad=0.)
plt.subplot(212, sharex=ax1)
plt.scatter(df.taille, df.time_val, label="Iteration de la valeur")
plt.scatter(df.taille, df.time_pol, label="Iteration de la politique")
plt.scatter(df.taille, df.time_pl, label="Programmation mathématique")
plt.xlabel("Taille de l'instance")
plt.ylabel("Durée d'execution")
plt.show()
def impact_p(interval_p,
nbr_tests=default_tests,
nbLignes=default_nbLignes,
nbCol=default_nbCol,
gamma=default_gamma,
epsilon=default_epsilon):
df = pd.DataFrame(columns=[
"p",
"iter_val",
"iter_pol",
"iter_pl",
"time_val",
"time_pol",
"time_pl",
])
for p in interval_p:
print("p = " + str(p))
somme1_it = 0
somme1_t = 0
somme2_it = 0
somme2_t = 0
somme3_it = 0
somme3_t = 0
for k in range(nbr_tests):
g = grille.Grille(nbLignes, nbCol, 2, 0.2, 0.2, 0.2, 0.2, 0.2,
[0, 1, 2, 3, 4], 1000).g
#iteration de la valeur
d, v, t, time = algos.iteration_de_la_valeur(g, p, gamma, epsilon)
somme1_it += t
somme1_t += time
#iteration de la politque
d, v, t, time = algos.iteration_de_la_politique(
g, p, gamma, epsilon)
somme2_it += t
somme2_t += time
#PL
d, v, t, time = algos.pl(g, p, gamma)
somme3_it += t
somme3_t += time
line = pd.Series({
'p': p,
'iter_val': somme1_it / nbr_tests,
'iter_pol': somme2_it / nbr_tests,
'iter_pl': somme3_it / nbr_tests,
'time_val': somme1_t / nbr_tests,
'time_pol': somme2_t / nbr_tests,
'time_pl': somme3_t / nbr_tests,
})
df = df.append(line, ignore_index=True)
print(df)
df.to_csv("./impact_p.csv", ';')
plt.figure(1)
ax1 = plt.subplot(211)
plt.title("Impact de p")
plt.scatter(df.p, df.iter_val, label="Iteration de la valeur")
plt.scatter(df.p, df.iter_pol, label="Iteration de la politique")
plt.scatter(df.p, df.iter_pl, label="Programmation mathématique")
plt.ylabel("Nombre d'itérations")
plt.setp(ax1.get_xticklabels(), visible=False)
plt.legend(bbox_to_anchor=(1, 1), loc=0, borderaxespad=0.)
plt.subplot(212, sharex=ax1)
plt.scatter(df.p, df.time_val, label="Iteration de la valeur")
plt.scatter(df.p, df.time_pol, label="Iteration de la politique")
plt.scatter(df.p, df.time_pl, label="Programmation mathématique")
plt.xlabel("Valeur de p")
plt.ylabel("Durée d'execution")
plt.show()