本文整理汇总了Python中bottleneck.argpartsort方法的典型用法代码示例。如果您正苦于以下问题：Python bottleneck.argpartsort方法的具体用法？Python bottleneck.argpartsort怎么用？Python bottleneck.argpartsort使用的例子？那么, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类bottleneck的用法示例。
在下文中一共展示了bottleneck.argpartsort方法的6个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: recall_at_multiple_ks_batch
# 需要导入模块: import bottleneck [as 别名]
# 或者: from bottleneck import argpartsort [as 别名]
def recall_at_multiple_ks_batch(train_data, heldout_data, Et, Eb, user_idx,
topks, vad_data):
batch_users = user_idx.stop - user_idx.start
X_pred = rec_eval._make_prediction(train_data, Et, Eb, user_idx,
batch_users, vad_data=vad_data)
recalls = np.empty((len(topks), batch_users))
for i, k in enumerate(topks):
idx = bn.argpartsort(-X_pred, k, axis=1)
X_pred_binary = np.zeros_like(X_pred, dtype=bool)
X_pred_binary[np.arange(batch_users)[:, np.newaxis], idx[:, :k]] = True
X_true_binary = (heldout_data[user_idx] > 0).toarray()
tmp = (np.logical_and(X_true_binary,
X_pred_binary).sum(axis=1)).astype(np.float32)
recalls[i] = tmp / X_true_binary.sum(axis=1)
return recalls 示例2: precision_at_k_batch
# 需要导入模块: import bottleneck [as 别名]
# 或者: from bottleneck import argpartsort [as 别名]
def precision_at_k_batch(train_data, heldout_data, Et, Eb, user_idx,
k=20, normalize=True, vad_data=None):
batch_users = user_idx.stop - user_idx.start
X_pred = _make_prediction(train_data, Et, Eb, user_idx,
batch_users, vad_data=vad_data)
idx = bn.argpartsort(-X_pred, k, axis=1)
X_pred_binary = np.zeros_like(X_pred, dtype=bool)
X_pred_binary[np.tile(np.arange(batch_users), (k, 1)).T, idx[:, :k]] = True
X_true_binary = (heldout_data[user_idx] > 0).toarray()
tmp = (np.logical_and(X_true_binary, X_pred_binary).sum(axis=1)).astype(
np.float32)
if normalize:
precision = tmp / np.minimum(k, X_true_binary.sum(axis=1))
else:
precision = tmp / k
return precision 示例3: recall_at_k_batch
# 需要导入模块: import bottleneck [as 别名]
# 或者: from bottleneck import argpartsort [as 别名]
def recall_at_k_batch(train_data, heldout_data, Et, Eb, user_idx,
k=20, vad_data=None):
batch_users = user_idx.stop - user_idx.start
X_pred = _make_prediction(train_data, Et, Eb, user_idx,
batch_users, vad_data=vad_data)
idx = bn.argpartsort(-X_pred, k, axis=1)
X_pred_binary = np.zeros_like(X_pred, dtype=bool)
X_pred_binary[np.tile(np.arange(batch_users), (k, 1)).T, idx[:, :k]] = True
X_true_binary = (heldout_data[user_idx] > 0).toarray()
tmp = (np.logical_and(X_true_binary, X_pred_binary).sum(axis=1)).astype(
np.float32)
recall = tmp / X_true_binary.sum(axis=1)
return recall 示例4: argsort
# 需要导入模块: import bottleneck [as 别名]
# 或者: from bottleneck import argpartsort [as 别名]
def argsort(x, topn=None):
"""Return indices of the `topn` greatest elements in numpy array `x`, in order."""
if topn is None:
topn = x.size
if topn <= 0:
return []
if topn >= x.size:
return numpy.argsort(x)[::-1]
biggest = bottleneck.argpartsort(x, x.size - topn)[-topn:]
# the indices in `biggest` are not sorted by magnitude => sort & return
return biggest.take(numpy.argsort(x.take(biggest))[::-1]) 示例5: argpartition
# 需要导入模块: import bottleneck [as 别名]
# 或者: from bottleneck import argpartsort [as 别名]
def argpartition(a, kth, axis=-1):
return bottleneck.argpartsort(a, kth, axis) 示例6: batch_gensample_model
# 需要导入模块: import bottleneck [as 别名]
# 或者: from bottleneck import argpartsort [as 别名]
def batch_gensample_model(in_queue, out_queue, f_init, f_next, k, gpulock, n_resample, minlen, maxlen, bleulowerbound = 0.25):
while True:
req = in_queue.get()
if req is None:
break
x, x_mask, tosample_seqs_x, tosample_seqs_y, fixed_seqs_x, fixed_seqs_y = preprocess_data(req[1], req[2], minlen, maxlen)
XX, YY = fixed_seqs_x, fixed_seqs_y
assert len(XX) == len(YY), 'len(XX) NE len(YY)'
WW = [1.] * len(XX)
resample_counter = 0
if x is not None:
samples, scores = beam_search_sample(f_init, f_next, x, x_mask, gpulock, k, maxlen)
resample_size = len(tosample_seqs_x)
for idx in xrange(resample_size):
bleu_scores = []
str_tt = [str(y) for y in tosample_seqs_y]
for ii in xrange(len(samples[idx])):
str_ss = [str(x) for x in samples[idx][ii]]
bleu_scores.append( nt.bleu([str_tt], str_ss, [0.25] * 4) )
sorted_id = bottleneck.argpartsort(-numpy.array(bleu_scores, dtype = 'float32'), n = n_resample)[:n_resample]
start_idx = 0
if bleu_scores[sorted_id[0]] < 0.999:
XX.append(tosample_seqs_x[idx])
YY.append(tosample_seqs_y[idx])
WW.append(1.)
start_idx = 1
for ii in xrange(start_idx, n_resample):
if bleu_scores[sorted_id[ii]] < bleulowerbound:
break
XX.append(tosample_seqs_x[idx])
YY.append(samples[idx][sorted_id[ii]])
WW.append(bleu_scores[sorted_id[ii]])
resample_counter += 1
out_queue.put((req[0], XX, YY, WW, resample_counter))
out_queue.put(None)
print 'Process %d Finish One' % os.getpid()
return