def al_sample_clusters(self, num_clusters, init):
'''
active learning with clusters
'''
strategy = self.strategy if not init else 'random'
sample_ids = []
for i in range(self.num_clusters):
if strategy == 'random':
new_sample_ids = rand_sample(self.unseen_ids[i], num_clusters)
for song_id in new_sample_ids:
self.unseen_ids[i].remove(song_id)
sample_ids += new_sample_ids
elif strategy == 'uncertainty':
unseen_x = get_x(self.train_data, ids=self.unseen_ids[i])
unseen_x_scores = self.learner.decision_function(
self.transform(unseen_x))
# get index of smallest absolute value (probability or
# distance from decision boundary)
# from unseen_x_scores, pop the corresponding entry from
# self.unseen_ids
sample_ids.append(self.unseen_ids[i].pop(
np.argmin(np.abs(unseen_x_scores))))
self.train_ids.update(sample_ids)
return sample_ids
def G1DConnBiasedMutateWeights(genome, **args):
## args['mutation_conns'] takes the list of indices of connections to be mutated
if not args['mutation_conns']:
mutation_conn_indices=rand_sample(numpy.array(len(genome)-1),int(args['pmut']*len(genome)))
# print "Indices of mutation weights:"
# print mutation_conn_indices
else:
mutation_conn_indices = args['mutation_conns']
# print "Indices of mutation weights:"
# print mutation_conn_indices
mu = genome.getParam("gauss_mu")
sigma = genome.getParam("gauss_sigma")
new_mutation_list=[]
for it in mutation_conn_indices:
final_value = genome['weight'][it] + rand_gauss(mu, sigma)
#final_value = min(final_value, genome.getParam("rangemax", Consts.CdefRangeMax))
#final_value = max(final_value, genome.getParam("rangemin", Consts.CdefRangeMin))
new_mutation_list.append(final_value)
numpy.put(genome['weight'],[mutation_conn_indices],[new_mutation_list])
return genome
def __get_used_samples(self):
indexes = rand_sample(range(0, len(self.samples)),
randint(1, self.max_data_in_combined))
used_samples = [k for i, k in enumerate(self.samples) if i in indexes]
fully_used = [
self.usage[k] == self.max_data_usage for k in used_samples
]
while all(fully_used):
indexes = rand_sample(range(0, len(self.samples)),
randint(1, self.max_data_in_combined))
used_samples = [
k for i, k in enumerate(self.samples) if i in indexes
]
fully_used = [self.usage[k] == 20 for k in used_samples]
return used_samples
def load_usps(filename, sample=-1):
with open(filename, "r") as f:
f.readline()
data = [[float(x) for x in l.split()] for l in f if len(l.split()) > 2]
tmp = np.array(data)
if sample > 0:
idx = rand_sample(range(len(tmp)), sample)
tmp = tmp[idx, :]
return tmp[:, 1:], tmp[:, 0].astype(int)
def spl(self,n,timing=False,nopref=False):
if timing: tic = time.time()
# Generate variates.
if nopref:
X=rand_sample(self.keys.values(),n)
else:
X =[]
for nn in xrange(n):
X.append(self.draw())
if timing: print('time elapsed:'+str(time.time()-tic))
return X
def al_sample(self, number, init):
'''
active learning sample
'''
strategy = self.strategy if not init else 'random'
if strategy == 'random':
# Initial samples must contain all labels for offline learners.
# This ensures that happens.
if init and not self.is_online():
sampled_classes = set()
while len(sampled_classes) != len(CLASSES):
sampled_classes = set()
sample_ids = rand_sample(self.unseen_ids, number)
sampled_classes = set()
for id_ in sample_ids:
sampled_classes.add(self.train_data[id_]['label'])
else:
sample_ids = rand_sample(self.unseen_ids, number)
for song_id in sample_ids:
self.unseen_ids.remove(song_id)
elif strategy == 'uncertainty':
unseen_x = get_x(self.train_data, ids=self.unseen_ids)
unseen_x_scores = self.learner.decision_function(
self.transform(unseen_x))
# get index of smallest absolute value (probability or distance from decision boundary)
# from unseen_x_scores, pop the corresponding entry from
# self.unseen_ids
sample_ids = [
self.unseen_ids.pop(np.argmin(np.abs(unseen_x_scores)))
]
self.train_ids.update(sample_ids)
return sample_ids
def split_data(data, proportion):
'''
Splits data into training and validation sets for simple classification.
:param data: complete labeled data
:param p: proportion of data to use for validation
:return train_data, validation_data
'''
validation_ids = rand_sample(list(data), int(len(data) * proportion))
validation_data = {k: data[k] for k in validation_ids}
train_data = {k: v for k, v in data.items() if k not in validation_data}
return FrozenMap(train_data), FrozenMap(validation_data)
def G1DConnUnbiasedMutateWeights(genome, **args):
## args['mutation_conns'] takes the list of indices of connections to be mutated
if not args['mutation_conns']:
mutation_conn_indices=rand_sample(numpy.arange(len(genome)),int(args['pmut']*len(genome)))
# print "Indices of mutation weights:"
# print mutation_conn_indices
else:
mutation_conn_indices = args['mutation_conns']
# print "Indices of mutation weights:"
# print mutation_conn_indices
new_mutation_list=[]
for it in mutation_conn_indices:
new_mutation_list.append(rand_uniform(genome.getParam("rangemin", Consts.CDefRangeMin),genome.getParam("rangemax", Consts.CDefRangeMax)))
numpy.put(genome['weight'],mutation_conn_indices,new_mutation_list)
return genome
def G1DConnUnbiasedMutateWeights(genome, **args):
## args['mutation_conns'] takes the list of indices of connections to be mutated
if not args.get('mutation_conns',0):
mutation_conn_indices=rand_sample(numpy.arange(len(genome)),int(args['pmut']*len(genome)))
# print "Indices of mutation weights:"
# print mutation_conn_indices
else:
mutation_conn_indices = args['mutation_conns']
# print "Indices of mutation weights:"
# print mutation_conn_indices
mu = genome.getParam("gauss_mu",0)
sigma = genome.getParam("gauss_sigma",1)
#new_mutation_list=[]
for it in mutation_conn_indices:
final_value=rand_gauss(mu, sigma)
genome['weight'][it] = final_value
# new_mutation_list.append(rand_uniform(genome.getParam("rangemin", Consts.CDefRangeMin),genome.getParam("rangemax", Consts.CDefRangeMax)))
mutations = len(mutation_conn_indices)
#numpy.put(genome['weight'],mutation_conn_indices,new_mutation_list)
return int(mutations)
def G1DConnMutateNodes(genome, **args): ## args['mutation_nodes'] takes the list of node numbers to be mutated if not args['mutation_nodes']: mutations=int(args['pmut']*len(numpy.unique(genome['to']))) mutation_node_numbers = rand_sample(numpy.unique(genome['to']), mutations) # print "randomly chosen mutation nodes:" # print mutation_node_numbers else: mutation_node_numbers = list(args['mutation_nodes']) # print "selected mutation nodes:" # print mutation_node_numbers mutations = len(arg['mutation_nodes']) for it in mutation_node_numbers: mutation_nodes = genome.getnode(it) # print "connections of mutation nodes" # print mutation_nodes mutation_conn_indices=getMutateConnIndex(genome, mutation_nodes) G1DConnBiasedMutateWeights(genome, mutation_conns=mutation_conn_indices) return int(mutations)
def ChromoInitializator(genome, **args):
genome.genomeList = rand_sample(xrange(genome.getListSize()), genome.getListSize())
def ChromoInitializator(genome, **args):
genome.genomeList = rand_sample(xrange(genome.getListSize()),
genome.getListSize())
def rand_max(arr, key=lambda x: x):
return rand_sample(max_els(arr, key), 1)[0]