def test_tlbo_variance():
lb = np.zeros(5)
lb[::2] = 1.1
ub = np.ones(5)
ub[::2] = 2.2
o = TLBO(f=lambda x: np.abs(np.mean(x) - 0.5),
lower_bound=lb,
upper_bound=ub)
o = helper_n_generations(o, 20)
print("costs history", o.bestcosts_)
solution, fitness = o.best()
print("best fitness", fitness[0])
print("best solution", solution[0])
assert len(o.fitness_) == 50
assert len(o.population_) == 50
assert len(o.bestcosts_) == 21
assert len(solution) == 1
assert len(fitness) == 1
def test_tlbo_constrained_himmelblau():
# function to minimize
def f(x):
x1, x2 = x
return ((x1**2) + x2 - 11)**2 + (x1 + (x2**2) - 7)**2
# constraint 2
def g1(x):
x1, x2 = x
return 26 - (x1 - 5)**2 - x2**2
# constraint 1 g2(x) >= 0
def g2(x):
x1, x2 = x
return 20 - (4 * x1) - x2
# penalty:
# 10 * v^2, if v < 0
# 0, otherwise
def p(v):
return 10 * min(v, 0)**2
# f'(x) - add penalties for constraints
def fp(x):
return f(x) + p(g1(x)) + p(g2(x))
lower_bounds = np.ones(2) * -5
upper_bounds = np.ones(2) * 5
tlbo = TLBO(fp, lower_bounds, upper_bounds, n_population=5)
tlbo = helper_n_generations(tlbo, 100)
best_solution, best_fitness = tlbo.best()
print("TLBO solution", best_solution, "fitness", best_fitness)
def build_for_class(self, X):
distance_fitness = lambda P: self.distance_sum(P, X)
# setup GA
ga = GeneticAlgorithm(fitness_function=distance_fitness,
elitism=3,
n_chromosomes=100,
n_genes=X.shape[1],
p_mutation=0.3)
ga = helper_n_generations(ga, self.n_iterations) # advance the GA
# return the best found parameters for this class.
chromosomes, fitness = ga.best(1)
return chromosomes[0]
def build_for_class(self, X):
distance_fitness = lambda P: self.distance_sum(P, X)
# setup GA
ga = GeneticAlgorithm(fitness_function=distance_fitness,
elitism=3,
n_chromosomes=100,
n_genes=X.shape[1],
p_mutation=0.3)
ga = helper_n_generations(ga, self.n_iterations) # advance the GA
# return the best found parameters for this class.
chromosomes, fitness = ga.best(1)
return chromosomes[0]
def build_for_class(self, X, y, class_idx):
# take column-wise min/mean/max for class
mins = np.nanmin(X[class_idx], 0)
means = np.nanmean(X[class_idx], 0)
maxs = np.nanmax(X[class_idx], 0)
ds = (maxs - mins) / 2.0
n_genes = 2 * self.m # adjustment for r and shrinking/expanding value for p/q
B = np.ones(n_genes)
def decode_with_shrinking_expanding(C):
def dcenter(j):
return min(1.0, max(0.0,
C[j])) - 0.5 if self.adjust_center else 1.0
return [
fl.PiSet(r=means[j] * dcenter(j),
p=means[j] - (ds[j] * C[j + 1]),
q=means[j] + (ds[j] * C[j + 1]))
for j in range(self.m)
]
y_target = np.zeros(y.shape) # create the target of 1 and 0.
y_target[class_idx] = 1.0
def rmse_fitness_function(chromosome):
proto = decode_with_shrinking_expanding(chromosome)
y_pred = _predict_one(proto, self.aggregation, X)
return mean_squared_error(y_target, y_pred)
logger.info("initializing GA %d iterations" % (self.iterations, ))
# initialize
ga = UnitIntervalGeneticAlgorithm(
fitness_function=helper_fitness(rmse_fitness_function),
crossover_function=UniformCrossover(0.5),
elitism=3,
n_chromosomes=100,
n_genes=n_genes,
p_mutation=0.3)
ga = helper_n_generations(ga, self.iterations)
chromosomes, fitnesses = ga.best(1)
return decode_with_shrinking_expanding(
chromosomes[0]), decode_with_shrinking_expanding(B)
def build_for_class(self, rs, X):
def distance_fitness(c):
return np.sum(np.abs(X - c))
# setup GA
ga = GeneticAlgorithm(fitness_function=helper_fitness(distance_fitness),
elitism=3,
n_chromosomes=100,
n_genes=X.shape[1],
p_mutation=0.3,
random_state=rs)
ga = helper_n_generations(ga, self.n_iterations) # advance the GA
# return the best found parameters for this class.
chromosomes, fitness = ga.best(1)
return chromosomes[0]
def build_for_class(self, rs, X):
def distance_fitness(c):
return np.sum(np.abs(X - c))
# setup GA
ga = GeneticAlgorithm(
fitness_function=helper_fitness(distance_fitness),
elitism=3,
n_chromosomes=100,
n_genes=X.shape[1],
p_mutation=0.3,
random_state=rs)
ga = helper_n_generations(ga, self.n_iterations) # advance the GA
# return the best found parameters for this class.
chromosomes, fitness = ga.best(1)
return chromosomes[0]
def test_jaya_sphere():
"""Example given in paper"""
o = JayaOptimizer(f=lambda x: np.sum(x**2),
lower_bound=np.ones(10) * -10.0,
upper_bound=np.ones(10) * 10.0,
n_population=34)
o = helper_n_generations(o, 100)
solution, fitness = o.best()
print("costs history", o.bestcosts_)
print("best fitness", fitness)
print("best solution", solution)
assert len(o.fitness_) == 34
assert len(o.population_) == 34
assert len(o.bestcosts_) == 101
def test_jaya_sphere_bounds():
""" Another example with strange domain """
o = JayaOptimizer(f=lambda x: np.sum(x**2),
lower_bound=np.array([1, 0.001, 100]),
upper_bound=np.array([10, 0.2, 1000]),
n_population=34)
o = helper_n_generations(o, 100)
solution, fitness = o.best()
print("costs history", o.bestcosts_)
print("best fitness", fitness)
print("best solution", solution)
assert len(o.fitness_) == 34
assert len(o.population_) == 34
assert len(o.bestcosts_) == 101
def test_jaya_sphere():
"""Example given in paper"""
o = JayaOptimizer(f=lambda x: np.sum(x**2),
lower_bound=np.ones(10) * -10.0,
upper_bound=np.ones(10) * 10.0,
n_population=34)
o = helper_n_generations(o, 100)
solution, fitness = o.best()
print("costs history", o.bestcosts_)
print("best fitness", fitness)
print("best solution", solution)
assert len(o.fitness_) == 34
assert len(o.population_) == 34
assert len(o.bestcosts_) == 101
def test_jaya_sphere_bounds():
""" Another example with strange domain """
o = JayaOptimizer(f=lambda x: np.sum(x**2),
lower_bound=np.array([1, 0.001, 100]),
upper_bound=np.array([10, 0.2, 1000]),
n_population=34)
o = helper_n_generations(o, 100)
solution, fitness = o.best()
print("costs history", o.bestcosts_)
print("best fitness", fitness)
print("best solution", solution)
assert len(o.fitness_) == 34
assert len(o.population_) == 34
assert len(o.bestcosts_) == 101
def build_for_class(self, X, y, class_idx):
# take column-wise min/mean/max for class
mins = np.nanmin(X[class_idx], 0)
means = np.nanmean(X[class_idx], 0)
maxs = np.nanmax(X[class_idx], 0)
ds = (maxs - mins) / 2.0
n_genes = 2 * self.m # adjustment for r and shrinking/expanding value for p/q
B = np.ones(n_genes)
def decode_with_shrinking_expanding(C):
def dcenter(j):
return min(1.0, max(0.0, C[j])) - 0.5 if self.adjust_center else 1.0
return [ fl.PiSet(r=means[j] * dcenter(j),
p=means[j] - (ds[j] * C[j + 1]),
q=means[j] + (ds[j] * C[j + 1]))
for j in range(self.m) ]
y_target = np.zeros(y.shape) # create the target of 1 and 0.
y_target[class_idx] = 1.0
def rmse_fitness_function(chromosome):
proto = decode_with_shrinking_expanding(chromosome)
y_pred = _predict_one(proto, self.aggregation, X)
return mean_squared_error(y_target, y_pred)
logger.info("initializing GA %d iterations" % (self.iterations,))
# initialize
ga = UnitIntervalGeneticAlgorithm(fitness_function=helper_fitness(rmse_fitness_function),
crossover_function=UniformCrossover(0.5),
elitism=3,
n_chromosomes=100,
n_genes=n_genes,
p_mutation=0.3)
ga = helper_n_generations(ga, self.iterations)
chromosomes, fitnesses = ga.best(1)
return decode_with_shrinking_expanding(chromosomes[0]), decode_with_shrinking_expanding(B)
def test_jaya_variance():
lb = np.zeros(5)
lb[::2] = 1.1
ub = np.ones(5)
ub[::2] = 2.2
o = JayaOptimizer(f=lambda x: np.abs(np.mean(x) - 0.5), lower_bound=lb, upper_bound=ub)
o = helper_n_generations(o, 20)
print("costs history", o.bestcosts_)
solution, fitness = o.best()
print("best fitness", fitness)
print("best solution", solution)
assert len(o.fitness_) == 50
assert len(o.population_) == 50
assert len(o.bestcosts_) == 21
def test_tlbo_alan_2():
def fitness(X):
print("X", X, "len", len(X))
x1, x2, x3, x4 = X.tolist()
return 0.0358 + (0.7349 * x1) + (.0578 * x2) - (0.3151 * x3) + (
0.6888 * x4) + (0.1803 * (x1**2)) - (0.0481 * (x2**2)) + (
0.1699 *
(x3**2)) - (0.0494 * (x4**2)) - (0.3555 * (x1 * x2)) - (
0.6316 * (x1 * x3)) + (0.5973 * (x1 * x4)) + (
0.0826 * (x2 * x3)) - (0.4736 *
(x2 * x4)) - (0.6547 *
(x3 * x4))
lower_bounds = np.array([-10, -10, -10, -10])
upper_bounds = np.array([10, 10, 10, 10])
tlbo = TeachingLearningBasedOptimizer(fitness, lower_bounds, upper_bounds)
tlbo = helper_n_generations(tlbo, 100)
best_solution, best_fitness = tlbo.best()
print("TLBO solution", best_solution, "fitness", best_fitness)
def test_tlbo_sphere():
"""Example given in matlab code"""
o = TeachingLearningBasedOptimizer(f=lambda x: np.sum(x**2),
lower_bound=np.ones(10) * -10.0,
upper_bound=np.ones(10) * 10.0,
n_population=34)
o = helper_n_generations(o, 50)
solution, fitness = o.best(3)
print("costs history", o.bestcosts_)
print("best fitness", fitness[0])
print("best solution", solution[0])
assert len(o.fitness_) == 34
assert len(o.population_) == 34
assert len(solution) == 3
assert len(fitness) == 3
def fit_weights(self, rs, models, X, y):
n_genes = self.n_models * len(self.classes_)
def fitness_function(c):
M = self.predict_(X, models, c)
y_pred = np.argmin(M, 1)
return 1.0 - accuracy_score(y, y_pred)
ga = GeneticAlgorithm(fitness_function=helper_fitness(fitness_function),
elitism=3,
n_chromosomes=100,
n_genes=n_genes,
p_mutation=0.3,
random_state=rs)
ga = helper_n_generations(ga, self.n_iterations_weights) # advance the GA
chromosomes, fitness = ga.best(1)
return chromosomes[0]
def test_tlbo_sphere():
"""Example given in matlab code"""
o = TeachingLearningBasedOptimizer(f=lambda x: np.sum(x**2),
lower_bound=np.ones(10) * -10.0,
upper_bound=np.ones(10) * 10.0,
n_population=34)
o = helper_n_generations(o, 50)
solution, fitness = o.best(3)
print("costs history", o.bestcosts_)
print("best fitness", fitness[0])
print("best solution", solution[0])
assert len(o.fitness_) == 34
assert len(o.population_) == 34
assert len(solution) == 3
assert len(fitness) == 3
def fit_weights(self, rs, models, X, y):
n_genes = self.n_models * len(self.classes_)
def fitness_function(c):
M = self.predict_(X, models, c)
y_pred = np.argmin(M, 1)
return 1.0 - accuracy_score(y, y_pred)
ga = GeneticAlgorithm(
fitness_function=helper_fitness(fitness_function),
elitism=3,
n_chromosomes=100,
n_genes=n_genes,
p_mutation=0.3,
random_state=rs)
ga = helper_n_generations(ga,
self.n_iterations_weights) # advance the GA
chromosomes, fitness = ga.best(1)
return chromosomes[0]