def sample_with_error(label, error_distribution, null_distribution):
"""Return label given error probability and null distributions
error_distribution must be of form {False: 1.0 - p_err, True: p_err}
"""
if discrete_sample(error_distribution):
# to generate error properly, draw from null distribution
return discrete_sample(null_distribution)
else:
# no error: append actual class label
return label
def sample_with_error(label, error_distribution, null_distribution):
"""Return label given error probability and null distributions
error_distribution must be of form {False: 1.0 - p_err, True: p_err}
"""
if discrete_sample(error_distribution):
# to generate error properly, draw from null distribution
return discrete_sample(null_distribution)
else:
# no error: append actual class label
return label
def simulate_predictions(n=100, seed=None):
"""simulate classifier predictions for data size of n
"""
if seed is None:
seed = random_seed()
np.random.seed(seed % (2 ** 32))
probas = np.random.random(n)
classes = [discrete_sample({0: (1 - p), 1: p}) for p in probas]
return classes, probas
def simulate_predictions(n=100, seed=None):
"""simulate classifier predictions for data size of n
"""
if seed is None:
seed = random_seed()
np.random.seed(seed % (2 ** 32))
probas = np.random.random(n)
classes = [discrete_sample({0: (1 - p), 1: p}) for p in probas]
return classes, probas
def mutate(self, seq):
"""
:param seq: sequence
:type seq: str
:returns: mutated sequence
:rtype: str
"""
delimiter = self.delimiter
doc_list = list(intersperse(delimiter, seq)) + [delimiter]
mutation_site = random.randint(0, len(doc_list) - 1)
from_letter = doc_list[mutation_site]
prob_dist = self.chain[from_letter]
to_letter = discrete_sample(prob_dist)
doc_list[mutation_site] = to_letter
return ''.join(el for el in doc_list if el != delimiter)
def generate(self, start, length):
"""Generate a sequence according to a Markov chain"""
for _ in xrange(length):
prob_dist = self.chain[start]
start = discrete_sample(prob_dist)
yield start
