def add_hypernyms(string: str, p: float = 0.01) -> str:
"""Replace word with a higher-level category.
A common negative sampling technique involves replacing words
in a sentence with a word that has the same general meaning, but
is too general for the context, e.g.:
"all dogs go to heaven" -> "all quadrupeds go to place"
The replacement words are drawn from wordnet (wordnet_). For
words with more than one possible replacement, one is selected using
``random.choice``.
Args:
string: text
p: conditional probability of replacing a word
Returns:
enriched text
.. _wordnet: https://wordnet.princeton.edu/
"""
wn = _get_wordnet()
words = [wn.lemmatize(w) for w in string.split()]
for index, word in enumerate(words):
if (word in HYPERNYMS) and random.binomial(1, p):
words[index] = random.choice(HYPERNYMS[word])
return " ".join(words)
def __init__(self, seqs, pred_dict, use_best_site=True):
"""
:param seqs: list[str]
:param pred_dict: dict
:return:
"""
super(GetScore, self).__init__()
width = len(next(pred_dict.iterkeys()))
self.ind = []
self.dists = []
for seq in seqs:
contigs = Subset(seq, width)
contigs.extend(RC(contigs))
scores = [pred_dict[key] for key in contigs] # len(key) == width
max_score = max(scores)
if use_best_site:
self.append(max_score)
else:
self.append(np.mean(scores))
ind = [i for i, x in enumerate(scores) if x == max_score]
idx = random.choice(ind)
self.ind.append(idx)
n = len(scores) / 2 # binding sites positions #
self.dists.append(idx - n / 2 if idx < n else idx - n - n / 2)
def add_synonyms(string: str, p: float = 0.01) -> str:
"""Replace word with one that has a close meaning.
A common data augmentation technique involves replacing words
in a sentence with a word that has the same general meaning
(arxiv:1509.01626_), e.g.:
"all dogs go to heaven" -> "all domestic dog depart to heaven"
The replacement words are drawn from wordnet (wordnet_). For
words with more than one possible replacement, one is selected using
``random.choice``.
Args:
string: text
p: conditional probability of replacing a word
Returns:
enriched text
.. _arxiv:1509.01626 : https://arxiv.org/abs/1509.01626
.. _wordnet: https://wordnet.princeton.edu/
"""
wn = _get_wordnet()
words = [wn.lemmatize(w) for w in string.split()]
for index, word in enumerate(words):
if (word in SYNONYMS) and random.binomial(1, p):
words[index] = random.choice(SYNONYMS[word])
return " ".join(words)
def sampling(self, x, len):
cell = self.create_cell()
sample = []
for t in range(len):
cell = self.forward_time(x, cell)
x = sprd.choice(range(self.vocab_size), p=cell[-1].y)
sample.append(x)
return sample
def sampling(self, x, len):
cell = self.create_cell()
sample = []
for t in range(len):
cell = self.forward_time(x, cell)
x = sprd.choice(range(self.vocab_size), p=cell[-1].y)
sample.append(x)
return sample
def draw(self):
"""
Draws the best arm with probability (1 - epsilon)
Draws any arm at random with probility epsilon
:return: The numerical index of the selected arm
"""
if random.rand() < self.epsilon:
return random.choice(self.n_arms)
else:
return argmax(self._metric_fn())
def draw(self):
"""
Draws the best arm with probability (1 - epsilon * temp)
Draws any arm with probability epsilon * temp
:return: The numerical index of the selected arm
"""
temp = self._schedule_fn(self.total_draws)
if random.rand() < self.epsilon * temp:
return random.choice(self.n_arms)
else:
return argmax(self._metric_fn())
def _forwardImplementation(self, inbuf, outbuf):
""" Draws a random number between 0 and 1. If the number is less
than epsilon, a random action is chosen. If it is equal or
larger than epsilon, the greedy action is returned.
"""
assert self.module
if random.random() < self.epsilon:
legalMoves = self.module.legalMoves.nonzero()[0]
outbuf[:] = array([random.choice(legalMoves)])
else:
outbuf[:] = inbuf
self.epsilon *= self.decay
def split_online_offline(dataset, online_ratio):
online_size = int(len(dataset) * online_ratio)
online_inds = random.choice(list(range(len(dataset))),
online_size,
replace=False)
online_mask = np.zeros(len(dataset))
online_mask[online_inds] = 1
offline_mask = np.ones(len(dataset))
offline_mask[online_inds] = 0.0
dataset = np.asarray(dataset)
online_dataset = dataset[online_mask == 1.0]
offline_dataset = dataset[offline_mask == 1.0]
return online_dataset, offline_dataset
def _forwardImplementation(self, inbuf, outbuf):
""" Draws a random number between 0 and 1. If the number is less
than epsilon, a random action is chosen. If it is equal or
larger than epsilon, the greedy action is returned.
"""
assert self.module
if random.random() < self.epsilon:
#only the actions that have a Q value > -infinity are valid
actionValues = self.module.getActionValues(self.state)
#print(actionValues)
actions = [a for a in xrange(len(actionValues)) if actionValues[a] > float("-inf")]
outbuf[:] = random.choice(actions)
else:
outbuf[:] = self.module.getMaxAction(self.state)
self.epsilon *= self.decay
def add_characters(string: str, p: float = 0.01) -> str:
"""Insert individual characters with probability p.
These are chosen randomly from the ascii alphabet (including
both upper and lower cases).
Args:
string: text
p: probability of removing a character
Returns:
enriched text
"""
for index in reversed(range(len(string))):
if random.binomial(1, p):
new_char = random.choice(list(ascii_letters))
string = string[:index] + new_char + string[index:]
return string
def sorted(self, size=None, step=1, reverse=True, rand=False):
"""
:param size: int | None
:param step: int
:param reverse: bool
:param rand: bool
:return: list[Peak]
"""
if rand:
out = random.choice(self, size=size, replace=False)
else:
sort_pks = sorted(self, reverse=reverse)
if size is None:
stop = None
else:
stop = step * size
assert stop <= len(self)
out = sort_pks[0:stop:step]
return Peaks(out)
def _sample_angles(self,N):
if (self.verbose): print " sample angles ..."
R = random.rand(N)
if self.ani:
# Anisotropic systems sample angles: cdf(q) = erfi(a*q)/erfi(a), a = sqrt(k)*p
p = self.r/self.ra
a = p*sqrt(self.k)
if self.mod.multi:
a *= sqrt(self.mod.s2/self.s2j)
self.q = numpy.zeros(N)
for j in range(N):
self.q[j] = optimize.brentq(self._pdf_angle, 0, 1, args=(a[j], R[j]))
else:
# Isotropic: cdf(q) = q
self.q = R
self.vr = self.v*self.q*random.choice((-1,1),size=self.N)
self.vt = self.v*sqrt(1-self.q**2)
def add_fat_thumbs(string: str, p: float = 0.01) -> str:
"""Replace characters with QWERTY neighbors.
One source of typographic mistakes comes from pressing a nearby key
on a keyboard (or on a touchscreen). With probability p, replace each
character is a string with one from a set of its neighbors. The
replacement is chosen using ``random.choice``.
Args:
string: text
p: probability of replacing a character
Returns:
enriched text
"""
for index, char in enumerate(string):
if char in NEIGHBORS and random.binomial(1, p):
new_char = random.choice(NEIGHBORS[char])
string = string[:index] + new_char + string[index + 1:]
return string
def _forwardImplementation(self, inbuf, outbuf):
""" Draws a random number between 0 and 1. If the number is less
than epsilon, a random action is chosen. If it is equal or
larger than epsilon, the greedy action is returned.
"""
assert self.module
if random.random() < self.epsilon:
#only the actions that have a Q value > -infinity are valid
actionValues = self.module.getActionValues(self.state)
#print(actionValues)
actions = [
a for a in xrange(len(actionValues))
if actionValues[a] > float("-inf")
]
outbuf[:] = random.choice(actions)
else:
outbuf[:] = self.module.getMaxAction(self.state)
self.epsilon *= self.decay
def _forwardImplementation(self, inbuf, outbuf):
""" Draws a random number between 0 and 1. If the number is less
than epsilon, a random action is chosen. If it is equal or
larger than epsilon, the greedy action is returned.
"""
assert self.module
# Choose action from allowed actions.
if random.random() < self.epsilon:
# Only select allowed actions
allowed_actions = np.where(
np.array(self.env.visited_states) == 0)[0]
act = array([random.choice(allowed_actions)])
outbuf[:] = act
else:
outbuf[:] = inbuf
self.epsilon *= self.decay
def _forwardImplementation(self, inbuf, outbuf):
""" Draws a random number between 0 and 1. If the number is less
than epsilon, a random action is chosen. If it is equal or
larger than epsilon, the greedy action is returned.
"""
assert self.module
#print np.nonzero(self.env.actions_index[self.env.current_state])[0]
# Choose action from allowed actions.
if random.random() < self.epsilon:
act = array(
[random.choice(self.env.mods[self.env.current_state]) - 1])
outbuf[:] = act
#print "current_state", self.env.current_state, "exploration action", act
#print "allowed actions", self.env.mods[self.env.current_state]
#outbuf[:] = array([random.randint(self.module.numActions)])
else:
outbuf[:] = inbuf
self.epsilon *= self.decay
def add_misspelling(string: str, p: float = 0.1) -> str:
"""Replace words with common misspellings.
Replaces a word with a common way that word is mispelled, given one or
more known, common misspellings taken from the Wikipedia spelling
correction corpus (wikipedia_). For words with more than one common
misspelling, one is chosen using ``random.choice``.
Args:
string: text
p: conditional probability of replacing a word
Returns:
enriched text
.. _wikipedia: https://en.wikipedia.org/wiki/Wikipedia:Lists_of_common_misspellings
"""
words = string.split()
for index, word in enumerate(words):
if (word in MISSPELLINGS) and random.binomial(1, p):
words[index] = random.choice(MISSPELLINGS[word])
return " ".join(words)
def _createorderArray(totalCount, selectedCount):
ordering = random.choice(list(range(1,int(totalCount+1))) , selectedCount)
return ordering
