def highPeakBestTriple(init_pop):
best_steerers = []
for t in range(len(limit_matrices)):
global_peak_index = convertGenotypeToInt(landscapes[t].getGlobalPeak())
best_prob = 1.0
best_s1 = -1
best_s2 = -1
best_s3 = -1
for s1 in range(len(limit_matrices)):
for s2 in range(len(limit_matrices)):
for s3 in range(len(limit_matrices)):
dist = np.array(init_pop * limit_matrices[s1] *
limit_matrices[s2] * limit_matrices[s3] *
limit_matrices[t])[0]
if dist[global_peak_index] < best_prob:
best_prob = dist[global_peak_index]
best_s1 = s1
best_s2 = s2
best_s3 = s3
print landscapes[t].name + ":", landscapes[
best_s1].name + " ---> " + landscapes[
best_s2].name + " ---> " + landscapes[
best_s3].name + ", ", best_prob
best_steerers.append((landscapes[best_s1], landscapes[best_s2],
landscapes[best_s3], best_prob))
return best_steerers
def lowestPeakBestTriple(init_pop):
best_steerers = []
for t in range(len(limit_matrices)):
lowest_peak_index = convertGenotypeToInt(
landscapes[t].getLowestFitnessPeak())
best_s1 = -1
best_s2 = -1
best_s3 = -1
best_prob = 0.0
for s1 in range(len(limit_matrices)):
for s2 in range(len(limit_matrices)):
for s3 in range(len(limit_matrices)):
dist = np.array(init_pop * limit_matrices[s1] *
limit_matrices[s2] * limit_matrices[s3] *
limit_matrices[t])[0]
if dist[lowest_peak_index] > best_prob:
best_prob = dist[lowest_peak_index]
best_s1 = s1
best_s2 = s2
best_s3 = s3
best_steerers.append((landscapes[best_s1], landscapes[best_s2],
landscapes[best_s3], best_prob))
print landscapes[t].name + ": ", landscapes[
best_s1].name + " ---> " + landscapes[
best_s2].name + " ---> " + landscapes[
best_s3].name + ", ", best_prob
return best_steerers
def doubleSteererPercentages(init_pop, allowSAM=True):
steering_stats = []
overall_better = 0
overall_worse = 0
overall_same = 0
for t in range(len(limit_matrices)):
if t != 10 or allowSAM:
better = 0
worse = 0
same = 0
peak_genotype = landscapes[t].getGlobalPeak()
peak_genotype_index = convertGenotypeToInt(peak_genotype)
peak_fitness = landscapes[t].getFitness(peak_genotype)
for k1 in range(len(limit_matrices)):
for k2 in range(len(limit_matrices)):
steered_pop = (init_pop * limit_matrices[k1] *
limit_matrices[k2] *
limit_matrices[t]).tolist()[0]
straight_pop = (init_pop * limit_matrices[t]).tolist()[0]
#Floating point arithematic is nasty, check for what is essential equality
if np.abs(steered_pop[peak_genotype_index] -
straight_pop[peak_genotype_index]) < 10**-15:
same += 1
elif steered_pop[peak_genotype_index] > straight_pop[
peak_genotype_index]:
worse += 1
elif steered_pop[peak_genotype_index] < straight_pop[
peak_genotype_index]:
better += 1
steering_stats.append((better, worse, same))
pc_better = 100 * float(better) / (better + worse + same)
pc_worse = 100 * float(worse) / (better + worse + same)
print landscapes[t].name + ", better:", str(better) + " (" + str(
pc_better) + "%)", ", worse:", str(worse) + " (" + str(
pc_worse) + "%)"
overall_better += better
overall_worse += worse
overall_same += same
pc_ov_better = 100 * float(overall_better) / (
overall_better + overall_worse + overall_same)
pc_ov_worse = 100 * float(overall_worse) / (
overall_better + overall_worse + overall_same)
print "Overall, better:", str(overall_better) + " (" + str(
pc_ov_better) + "%)", ", worse:", str(overall_worse) + " (" + str(
pc_ov_worse) + "%)"
return steering_stats
def probLowestPeak(init_pop):
lowest_pgs_probs = []
for t in range(len(limit_matrices)):
dist = np.array(init_pop * limit_matrices[t])[0]
lpg = landscapes[t].getLowestFitnessPeak()
lowest_pgs_probs.append(dist[convertGenotypeToInt(lpg)])
print landscapes[t].name + ":", lowest_pgs_probs[t]
return lowest_pgs_probs
def probHighestPeak(init_pop):
highest_pgs_probs = []
for t in range(len(limit_matrices)):
dist = np.array(init_pop * limit_matrices[t])[0]
global_peak_index = convertGenotypeToInt(landscapes[t].getGlobalPeak())
highest_pgs_probs.append(dist[global_peak_index])
print landscapes[t].name + ":", highest_pgs_probs[t]
return highest_pgs_probs
def probHighestPeak(init_pop): highest_pgs_probs = [] for t in range(len(limit_matrices)): dist = np.array(init_pop * limit_matrices[t])[0] global_peak_index = convertGenotypeToInt(landscapes[t].getGlobalPeak()) highest_pgs_probs.append(dist[global_peak_index]) print landscapes[t].name+":", highest_pgs_probs[t] return highest_pgs_probs
def probLowestPeak(init_pop): lowest_pgs_probs = [] for t in range(len(limit_matrices)): dist = np.array(init_pop * limit_matrices[t])[0] lpg = landscapes[t].getLowestFitnessPeak() lowest_pgs_probs.append(dist[convertGenotypeToInt(lpg)]) print landscapes[t].name+":", lowest_pgs_probs[t] return lowest_pgs_probs
def tripleSteererPercentages(init_pop, allowSAM = True):
steering_stats = []
overall_better = 0
overall_worse = 0
overall_same = 0
for t in range(len(limit_matrices)):
if t!=10 or allowSAM:
better = 0
worse = 0
same = 0
peak_genotype = landscapes[t].getGlobalPeak()
peak_genotype_index = convertGenotypeToInt(peak_genotype)
peak_fitness = landscapes[t].getFitness(peak_genotype)
for k1 in range(len(limit_matrices)):
for k2 in range(len(limit_matrices)):
for k3 in range(len(limit_matrices)):
steered_pop = (init_pop * limit_matrices[k1] * limit_matrices[k2] * limit_matrices[k3] * limit_matrices[t]).tolist()[0]
straight_pop = (init_pop * limit_matrices[t]).tolist()[0]
#Floating point arithematic is nasty, check for what is essential equality
if np.abs(steered_pop[peak_genotype_index] - straight_pop[peak_genotype_index]) < 10**-15:
same+=1
elif steered_pop[peak_genotype_index] > straight_pop[peak_genotype_index]:
worse+=1
elif steered_pop[peak_genotype_index] < straight_pop[peak_genotype_index]:
better+=1
steering_stats.append((better,worse,same))
pc_better = 100*float(better) / (better+worse+same)
pc_worse = 100*float(worse) / (better+worse+same)
print landscapes[t].name+", better:", str(better)+" ("+str(pc_better)+"%)", ", worse:", str(worse)+" ("+str(pc_worse)+"%)"
overall_better+=better
overall_worse+=worse
overall_same+=same
pc_ov_better = 100*float(overall_better) / (overall_better+overall_worse+overall_same)
pc_ov_worse = 100*float(overall_worse) / (overall_better+overall_worse+overall_same)
print "Overall, better:", str(overall_better)+" ("+str(pc_ov_better)+"%)", ", worse:", str(overall_worse)+" ("+str(pc_ov_worse)+"%)"
return steering_stats
def highPeakBestSingle(init_pop): best_steerers = [] for t in range(len(limit_matrices)): global_peak_index = convertGenotypeToInt(landscapes[t].getGlobalPeak()) best_prob = 1.0 best_steerer = -1 for s in range(len(limit_matrices)): dist = np.array(init_pop * limit_matrices[s] * limit_matrices[t])[0] if dist[global_peak_index] < best_prob: best_prob = dist[global_peak_index] best_steerer = s best_steerers.append((landscapes[s], best_prob)) print landscapes[t].name+":", landscapes[best_steerer].name, best_prob return best_steerers
def highPeakBestSingle(init_pop):
best_steerers = []
for t in range(len(limit_matrices)):
global_peak_index = convertGenotypeToInt(landscapes[t].getGlobalPeak())
best_prob = 1.0
best_steerer = -1
for s in range(len(limit_matrices)):
dist = np.array(init_pop * limit_matrices[s] *
limit_matrices[t])[0]
if dist[global_peak_index] < best_prob:
best_prob = dist[global_peak_index]
best_steerer = s
best_steerers.append((landscapes[s], best_prob))
print landscapes[t].name + ":", landscapes[
best_steerer].name, best_prob
return best_steerers
def highPeakBestDouble(init_pop): best_steerers = [] for t in range(len(limit_matrices)): global_peak_index = convertGenotypeToInt(landscapes[t].getGlobalPeak()) best_prob = 1.0 best_s1 = -1 best_s2 = -1 for s1 in range(len(limit_matrices)): for s2 in range(len(limit_matrices)): dist = np.array(init_pop * limit_matrices[s1] * limit_matrices[s2] * limit_matrices[t])[0] if dist[global_peak_index] < best_prob: best_prob = dist[global_peak_index] best_s1 = s1 best_s2 = s2 print landscapes[t].name+":", landscapes[best_s1].name+" ---> "+landscapes[best_s2].name+", ", best_prob best_steerers.append((landscapes[best_s2], landscapes[best_s2], best_prob)) return best_steerers
def lowestPeakBestDouble(init_pop): best_steerers = [] for t in range(len(limit_matrices)): lowest_peak_index = convertGenotypeToInt(landscapes[t].getLowestFitnessPeak()) best_s1 = -1 best_s2 = -1 best_prob = 0.0 for s1 in range(len(limit_matrices)): for s2 in range(len(limit_matrices)): dist = np.array(init_pop * limit_matrices[s1] * limit_matrices[s2] * limit_matrices[t])[0] if dist[lowest_peak_index] > best_prob: best_prob = dist[lowest_peak_index] best_s1 = s1 best_s2 = s2 best_steerers.append((landscapes[best_s1], landscapes[best_s2], best_prob)) print landscapes[t].name+": ", landscapes[best_s1].name+" ---> "+landscapes[best_s2].name+", ", best_prob return best_steerers
def lowestPeakBestSingle(init_pop): best_steerers = [] for t in range(len(limit_matrices)): lowest_peak_index = convertGenotypeToInt(landscapes[t].getLowestFitnessPeak()) best_steerer = -1 best_prob = 0.0 #For each possible steerer for s in range(len(limit_matrices)): dist = np.array(init_pop * limit_matrices[s] * limit_matrices[t])[0] if dist[lowest_peak_index] > best_prob: best_prob = dist[lowest_peak_index] best_steerer = s best_steerers.append((landscapes[best_steerer], best_prob)) print landscapes[t].name+":", best_steerers[t][0].name, best_steerers[t][1] return best_steerers
def lowestPeakBestSingle(init_pop):
best_steerers = []
for t in range(len(limit_matrices)):
lowest_peak_index = convertGenotypeToInt(
landscapes[t].getLowestFitnessPeak())
best_steerer = -1
best_prob = 0.0
#For each possible steerer
for s in range(len(limit_matrices)):
dist = np.array(init_pop * limit_matrices[s] *
limit_matrices[t])[0]
if dist[lowest_peak_index] > best_prob:
best_prob = dist[lowest_peak_index]
best_steerer = s
best_steerers.append((landscapes[best_steerer], best_prob))
print landscapes[t].name + ":", best_steerers[t][
0].name, best_steerers[t][1]
return best_steerers