def get(self, X):
X = np.array(X)
X_nan = np.isnan(X)
imputed = self.meanImput(X.copy())
if len(self.estimators_) > 1:
for i, estimator_ in enumerate(self.estimators_):
X_s = np.delete(imputed, i, 1)
y_nan = X_nan[:, i]
X_unk = X_s[y_nan]
result_ = []
if len(X_unk) > 0:
for unk in X_unk:
result_.append(estimator_.predict(unk))
X[y_nan, i] = result_
return X
python类delete()的实例源码
def loadLogoSet(path, rows,cols,test_data_rate=0.15):
random.seed(612)
_, imgID = readItems('data.txt')
y, _ = modelDict(path)
nPics = len(y)
faceassset = np.zeros((nPics,rows,cols), dtype = np.uint8) ### gray images
noImg = []
for i in range(nPics):
temp = cv2.imread(path +'logo/'+imgID[i]+'.jpg', 0)
if temp == None:
noImg.append(i)
elif temp.size < 1000:
noImg.append(i)
else:
temp = cv2.resize(temp,(cols, rows), interpolation = cv2.INTER_CUBIC)
faceassset[i,:,:] = temp
y = np.delete(y, noImg,0); faceassset = np.delete(faceassset, noImg, 0)
nPics = len(y)
index = random.sample(np.arange(nPics), int(nPics*test_data_rate))
x_test = faceassset[index,:,:]; x_train = np.delete(faceassset, index, 0)
y_test = y[index]; y_train = np.delete(y, index, 0)
return (x_train, y_train), (x_test, y_test)
def repeat(tensor: tf.Tensor, repeats: int, axis: int) -> tf.Tensor:
"""
Repeat elements of the input tensor in the specified axis ``repeats``-times.
.. note::
Chaining of this op may produce TF warnings although the performance seems to be unaffected.
:param tensor: TF tensor to be repeated
:param repeats: number of repeats
:param axis: axis to repeat
:return: tensor with repeated elements
"""
shape = tensor.get_shape().as_list()
dims = np.arange(len(tensor.shape))
prepare_perm = np.hstack(([axis], np.delete(dims, axis)))
restore_perm = np.hstack((dims[1:axis+1], [0], dims[axis+1:]))
indices = tf.cast(tf.floor(tf.range(0, shape[axis]*repeats)/tf.constant(repeats)), 'int32')
shuffled = tf.transpose(tensor, prepare_perm)
repeated = tf.gather(shuffled, indices)
return tf.transpose(repeated, restore_perm)
def main():
iris = load_iris()
test_idx = [0, 50, 100]
# training Data
train_target = np.delete(iris.target, test_idx)
train_data = np.delete(iris.data, test_idx, axis=0)
# testing data
test_target = iris.target[test_idx]
test_data = iris.data[test_idx]
# Train Classifier
clf = tree.DecisionTreeClassifier()
clf = clf.fit(train_data, train_target)
print(clf.predict(test_data))
# Run main
def _calc_B_for_tetra3d11(nodes,volume):
A = np.ones((4,4))
belta = np.zeros(4)
gama = np.zeros(4)
delta = np.zeros(4)
for i,nd in enumerate(nodes):
A[i,1:] = nd.coord
for i in range(4):
belta[i] = (-1)**(i+1)*np.linalg.det(np.delete(np.delete(A,i,0),1,1))
gama[i] = (-1)**(i+2)*np.linalg.det(np.delete(np.delete(A,i,0),2,1))
delta[i] = (-1)**(i+1)*np.linalg.det(np.delete(np.delete(A,i,0),3,1))
B = 1./(6.*volume)*np.array([[belta[0],0.,0.,belta[1],0.,0.,belta[2],0.,0.,belta[3],0.,0.],
[0.,gama[0],0.,0.,gama[1],0.,0.,gama[2],0.,0.,gama[3],0.],
[0.,0.,delta[0],0.,0.,delta[1],0.,0.,delta[2],0.,0.,delta[3]],
[gama[0],belta[0],0.,gama[1],belta[1],0.,gama[2],belta[2],0,gama[3],belta[3],0.],
[0.,delta[0],gama[0],0.,delta[1],gama[1],0.,delta[2],gama[2],0.,delta[3],gama[3]],
[delta[0],0.,belta[0],delta[1],0.,belta[1],delta[2],0.,belta[2],delta[3],0,belta[3]]])
return B
def append_neg_and_retrain(self, feat=None, force=False):
if feat is not None:
num = feat.shape[0]
self.neg = np.vstack((self.neg, feat))
self.num_neg_added += num
if self.num_neg_added > self.retrain_limit or force:
self.num_neg_added = 0
new_w_b, pos_scores, neg_scores = self.train()
# scores = np.dot(self.neg, new_w_b[0].T) + new_w_b[1]
# easy_inds = np.where(neg_scores < self.evict_thresh)[0]
not_easy_inds = np.where(neg_scores >= self.evict_thresh)[0]
if len(not_easy_inds) > 0:
self.neg = self.neg[not_easy_inds, :]
# self.neg = np.delete(self.neg, easy_inds)
print(' Pruning easy negatives')
print(' Cache holds {} pos examples and {} neg examples'.
format(self.pos.shape[0], self.neg.shape[0]))
print(' {} pos support vectors'.format((pos_scores <= 1).sum()))
print(' {} neg support vectors'.format((neg_scores >= -1).sum()))
return new_w_b
else:
return None
def tune_tal(mono_phi_score, tal_list):
errs = []
tals = []
for tal in tal_list:
err = []
for i in range(len(mono_phi_score)):
mono_1 = numpy.delete(mono_phi_score, i, axis=0)
dim_h = mono_phi_score[i][:-1]
value_h, alpha = train_predict_regression(mono_1, dim_h, tal)
err.append((value_h - mono_phi_score[i][-1])**2)
err = numpy.mean(err)
errs.append(err)
tals.append(tal)
print 'regression tal:', tal, 'err', err
idx = numpy.argmin(errs)
return tals[idx]
def append_neg_and_retrain(self, feat=None, force=False):
if feat is not None:
num = feat.shape[0]
self.neg = np.vstack((self.neg, feat))
self.num_neg_added += num
if self.num_neg_added > self.retrain_limit or force:
self.num_neg_added = 0
new_w_b, pos_scores, neg_scores = self.train()
# scores = np.dot(self.neg, new_w_b[0].T) + new_w_b[1]
# easy_inds = np.where(neg_scores < self.evict_thresh)[0]
not_easy_inds = np.where(neg_scores >= self.evict_thresh)[0]
if len(not_easy_inds) > 0:
self.neg = self.neg[not_easy_inds, :]
# self.neg = np.delete(self.neg, easy_inds)
print(' Pruning easy negatives')
print(' Cache holds {} pos examples and {} neg examples'.
format(self.pos.shape[0], self.neg.shape[0]))
print(' {} pos support vectors'.format((pos_scores <= 1).sum()))
print(' {} neg support vectors'.format((neg_scores >= -1).sum()))
return new_w_b
else:
return None
def append_neg_and_retrain(self, feat=None, force=False):
if feat is not None:
num = feat.shape[0]
self.neg = np.vstack((self.neg, feat))
self.num_neg_added += num
if self.num_neg_added > self.retrain_limit or force:
self.num_neg_added = 0
new_w_b, pos_scores, neg_scores = self.train()
# scores = np.dot(self.neg, new_w_b[0].T) + new_w_b[1]
# easy_inds = np.where(neg_scores < self.evict_thresh)[0]
not_easy_inds = np.where(neg_scores >= self.evict_thresh)[0]
if len(not_easy_inds) > 0:
self.neg = self.neg[not_easy_inds, :]
# self.neg = np.delete(self.neg, easy_inds)
print(' Pruning easy negatives')
print(' Cache holds {} pos examples and {} neg examples'.
format(self.pos.shape[0], self.neg.shape[0]))
print(' {} pos support vectors'.format((pos_scores <= 1).sum()))
print(' {} neg support vectors'.format((neg_scores >= -1).sum()))
return new_w_b
else:
return None
def eliminate_overlapping_locations(f, separation):
""" Makes sure that no position is within `separation` from each other, by
deleting one of the that are to close to each other.
"""
separation = validate_tuple(separation, f.shape[1])
assert np.greater(separation, 0).all()
# Rescale positions, so that pairs are identified below a distance of 1.
f = f / separation
while True:
duplicates = cKDTree(f, 30).query_pairs(1)
if len(duplicates) == 0:
break
to_drop = []
for pair in duplicates:
to_drop.append(pair[1])
f = np.delete(f, to_drop, 0)
return f * separation
def setdiff(eq1, eq2):
eq1, eq2 = eqsize(eq1, eq2)
c1 = [None] * eq1.shape
c2 = [None] * eq2.shape
for i in range(0, eq1.size):
c1.append[i] = hash(eq2[i])
for i in range(0, eq2.size):
c2[i] = hash(eq2[i])
ia = np.delete(np.arange(np.alen(c1)), np.searchsorted(c1, c2))
ia = (ia[:]).conj().T
p = eq1[ia]
return p, ia
def McCormack(x_nods_quantity, grid, transfer_velocity, time_step, x_step):
if (transfer_velocity[0] > 0):
new_grid = grid
for m in range(2, x_nods_quantity - 1):
sigma = transfer_velocity[m] * time_step / x_step
new_grid[m] = grid[m] - np.dot(sigma, (grid[m] - grid[m-1])) + \
np.dot(sigma**2, (grid[m] - grid[m-2]))
else:
new_grid = grid
for m in range(2, x_nods_quantity - 1):
sigma = transfer_velocity[m] * time_step / x_step
new_grid[m] = grid[m] - np.dot(sigma, (grid[m+1] - grid[m])) + \
np.dot(sigma ** 2, (grid[m+2] - grid[m]))
#new_grid = np.delete(grid, [0, 1])
# returning array without additional nod and border condition
return new_grid
def test_silence_frame_removal_given_hts_labels():
qs_file_name = join(DATA_DIR, "questions-radio_dnn_416.hed")
binary_dict, continuous_dict = hts.load_question_set(qs_file_name)
input_state_label = join(DATA_DIR, "label_state_align", "arctic_a0001.lab")
labels = hts.load(input_state_label)
features = fe.linguistic_features(labels,
binary_dict,
continuous_dict,
add_frame_features=True,
subphone_features="full"
)
# Remove silence frames
indices = labels.silence_frame_indices()
features = np.delete(features, indices, axis=0)
y = np.fromfile(join(DATA_DIR, "nn_no_silence_lab_425", "arctic_a0001.lab"),
dtype=np.float32).reshape(-1, features.shape[-1])
assert features.shape == y.shape
assert np.allclose(features, y)
# Make sure we can get same results with Merlin
def validate(self):
wav_dir = join(self.data_root, self.subset, "wav")
if not isdir(wav_dir):
raise RuntimeError("{} doesn't exist.".format(wav_dir))
miss_indices = []
for idx, name in enumerate(self.names):
wav_path = join(wav_dir, name + ".wav")
if not exists(wav_path):
miss_indices.append(idx)
if len(miss_indices) > 0:
warn("{}/{} wav files were missing in subset {}.".format(
len(miss_indices), len(self.names), self.subset))
self.names = np.delete(self.names, miss_indices)
self.transcriptions = np.delete(self.transcriptions, miss_indices)
def data_split(arr):
'''
num2 = df.values
num2 = np.delete(num2,)
'''
df2 = df
df3 = df
#print arr
df2 = df2.drop([i for i in arr])
df3 = df3.drop([i for i in xrange(0,len(df)) if i not in arr])
return (df2,df3)
def FileReader(file_list,param_list):
row_add = np.zeros(shape=(1,len(param_list)+1))
for file in file_list:
hdulist = fits.open(file,memmap=True)
data_in = hdulist[1].data
col_add = np.zeros(shape=(len(data_in),1))
print file
for param in param_list:
data_now = np.reshape(data_in[param],(len(data_in[param]),1))
col_add = np.append(col_add,data_now,axis=1)
row_add = np.append(row_add,col_add,axis=0)
del hdulist
row_add = np.delete(row_add,0,axis=0)
row_add = np.delete(row_add,0,axis=1)
return row_add
def create_vertex_groups(groups=['common', 'not_used'], weights=[0.0, 0.0], ob=None):
'''Creates vertex groups and sets weights. "groups" is a list of strings
for the names of the groups. "weights" is a list of weights corresponding
to the strings. Each vertex is assigned a weight for each vertex group to
avoid calling vertex weights that are not assigned. If the groups are
already present, the previous weights will be preserved. To reset weights
delete the created groups'''
if ob is None:
ob = bpy.context.object
vg = ob.vertex_groups
for g in range(0, len(groups)):
if groups[g] not in vg.keys(): # Don't create groups if there are already there
vg.new(groups[g])
vg[groups[g]].add(range(0,len(ob.data.vertices)), weights[g], 'REPLACE')
else:
vg[groups[g]].add(range(0,len(ob.data.vertices)), 0, 'ADD') # This way we avoid resetting the weights for existing groups.
def linregress(self):
"""Get the linear regression of the mean values in this plot. Returns
a tuple containing the best-fit line y-values for this plotter's
t_axis, the drift coefficient, and the ``linregress`` named tuple from
scipy.stats.linregress."""
cleandata = np.delete(self.plot_vars.means, self.bad_indices.means)
cleantimes = np.delete(self.t_axis, self.bad_indices.means)
if len(cleandata) != 0:
r = scipy.stats.linregress(cleantimes, cleandata)
bestfit = r.slope * self.t_axis + r.intercept
driftcoeff = r.slope / SEC_PER[self.t_units]
else:
bestfit = 0
driftcoeff = 0
r = None
return self.LinRegress(bestfit=bestfit, driftcoeff=driftcoeff,
linregress=r)
def trend(self):
"""Subtract the trend specified in
``Plotter.plot_properties['detrend']`` from each plot. Trend can be
the 'mean' value of the plot, the 'linear' least squares best fit, a
custom-specified number, or simply 'none' if no trend should be
removed."""
if self.plot_properties['detrend'] == 'mean':
# delete bad indices before calculating the trend, since they
# can skew the trend.
cleandata = np.delete(self.plot_vars.means, self.bad_indices.means)
if len(cleandata) != 0:
trend = cleandata.mean()
else:
trend = 0
elif self.plot_properties['detrend'] == 'none':
trend = 0
elif self.plot_properties['detrend'] == 'linear':
trend, driftcoeff, linregress = self.linregress
else:
trend = self.plot_properties['detrend']
return trend
def plot_timeseries(self, ax, **kwargs):
"""Scale up by 10^9 since plots are in ns, not seconds.
Remove any indices considered bad in ``plot_properties``"""
# define the variables for our plots
y = np.delete(self.plot_vars.means - self.trend,
self.bad_indices.means) / SEC_PER['ns']
t = np.delete(self.t_axis, self.bad_indices.means)
yerr = np.delete(self.plot_vars.stds,
self.bad_indices.means) / SEC_PER['ns']
mint = np.delete(self.t_axis, self.bad_indices.mins)
miny = np.delete(self.plot_vars.mins - self.trend,
self.bad_indices.mins) / SEC_PER['ns']
maxt = np.delete(self.t_axis, self.bad_indices.maxs)
maxy = np.delete(self.plot_vars.maxs - self.trend,
self.bad_indices.maxs) / SEC_PER['ns']
# plot everything, but only if the plotted data has nonzero length
# in order to avoid an annoying matplotlib bug when adding legends.
if len(t) != 0:
ax.errorbar(t, y, marker="o", color="green", linestyle='none',
yerr=yerr, label="Means +/- Std. Dev.")
if len(mint) != 0:
ax.scatter(mint, miny, marker="^", color="blue", label="Minima")
if len(maxt) != 0:
ax.scatter(maxt, maxy, marker="v", color="red", label="Maxima")
def plot_timeseries(self, ax, **kwargs):
"""Scale up by 10^9 since plots are in ns, not seconds.
Remove any indices considered bad in ``plot_properties``"""
# define the variables for our plots
t = np.delete(self.t_axis, self.bad_indices.means)
y = np.delete(self.plot_vars.means - self.trend,
self.bad_indices.means) / SEC_PER['ns']
yerr = np.delete(self.plot_vars.stds,
self.bad_indices.means) / SEC_PER['ns']
mint = np.delete(self.t_axis, self.bad_indices.absmins)
miny = np.delete(self.plot_vars.absmins - self.trend,
self.bad_indices.absmins) / SEC_PER['ns']
maxt = np.delete(self.t_axis, self.bad_indices.absmaxs)
maxy = np.delete(self.plot_vars.absmaxs - self.trend,
self.bad_indices.absmaxs) / SEC_PER['ns']
# plot everything, but only if the plotted data has nonzero length
# in order to avoid an annoying matplotlib bug when adding legends.
if len(t) != 0:
ax.errorbar(t, y, marker="o", color="green", linestyle='none',
yerr=yerr, label="Means +/- Std. Dev.")
if len(mint) != 0:
ax.scatter(mint,miny,marker="^", color="blue", label="Abs. Minima")
if len(maxt) != 0:
ax.scatter(maxt,maxy,marker="v", color="red", label="Abs. Maxima")
def plot_timeseries(self, ax, **kwargs):
ax.plot(np.delete(self.t_axis, self.bad_indices.means),
np.delete(self.plot_vars.means - self.trend,
self.bad_indices.means) / SEC_PER['ns'],
marker="o", color="green", label="Recorded Signal")
# put the start and/or end time in the plot as a vertical line
unitfactor = SEC_PER[self.t_units]
dq_start = (self.dq_segment.start.gpsSeconds - self.start) / unitfactor
dq_end = (self.dq_segment.end.gpsSeconds - self.start) / unitfactor
zorder = self.plot_properties['start_end_zorder']
if self.t_lim[0] <= dq_start:
deep_pink = '#FF1493'
plot_vertical_marker(ax, [dq_start], zorder=zorder,
label="Start of Segment", color=deep_pink)
if dq_end <= self.t_lim[1]:
midnight_blue = '#191970'
plot_vertical_marker(ax, [dq_end], zorder=zorder,
label="End of Segment", color=midnight_blue)
def append_neg_and_retrain(self, feat=None, force=False):
if feat is not None:
num = feat.shape[0]
self.neg = np.vstack((self.neg, feat))
self.num_neg_added += num
if self.num_neg_added > self.retrain_limit or force:
self.num_neg_added = 0
new_w_b, pos_scores, neg_scores = self.train()
# scores = np.dot(self.neg, new_w_b[0].T) + new_w_b[1]
# easy_inds = np.where(neg_scores < self.evict_thresh)[0]
not_easy_inds = np.where(neg_scores >= self.evict_thresh)[0]
if len(not_easy_inds) > 0:
self.neg = self.neg[not_easy_inds, :]
# self.neg = np.delete(self.neg, easy_inds)
print(' Pruning easy negatives')
print(' Cache holds {} pos examples and {} neg examples'.
format(self.pos.shape[0], self.neg.shape[0]))
print(' {} pos support vectors'.format((pos_scores <= 1).sum()))
print(' {} neg support vectors'.format((neg_scores >= -1).sum()))
return new_w_b
else:
return None
def remove_indexes(self, rm_idx_list, rearranged_props):
"""
The k-points with velocity < 1 cm/s (either in valence or conduction band) are taken out as those are
troublesome later with extreme values (e.g. too high elastic scattering rates)
:param rm_idx_list ([int]): the kpoint indexes that need to be removed for each property
:param rearranged_props ([str]): list of properties for which some indexes need to be removed
:return:
"""
for i, tp in enumerate(["n", "p"]):
for ib in range(self.cbm_vbm[tp]["included"]):
rm_idx_list_ib = list(set(rm_idx_list[tp][ib]))
rm_idx_list_ib.sort(reverse=True)
rm_idx_list[tp][ib] = rm_idx_list_ib
logging.debug("# of {}-type kpoints indexes with low velocity or off-energy: {}".format(tp,len(rm_idx_list_ib)))
for prop in rearranged_props:
self.kgrid[tp][prop] = np.array([np.delete(self.kgrid[tp][prop][ib], rm_idx_list[tp][ib], axis=0) \
for ib in range(self.cbm_vbm[tp]["included"])])
def transform(self, X):
check_is_fitted(self, ['statistics_', 'estimators_', 'gamma_'])
X = check_array(X, copy=True, dtype=np.float64, force_all_finite=False)
if X.shape[1] != self.statistics_.shape[1]:
raise ValueError("X has %d features per sample, expected %d"
% (X.shape[1], self.statistics_.shape[1]))
X_nan = np.isnan(X)
imputed = self.initial_imputer.fit_transform(X)
if len(self.estimators_) > 1:
for i, estimator_ in enumerate(self.estimators_):
X_s = np.delete(imputed, i, 1)
y_nan = X_nan[:, i]
X_unk = X_s[y_nan]
if len(X_unk) > 0:
X[y_nan, i] = estimator_.predict(X_unk)
else:
estimator_ = self.estimators_[0]
X[X_nan] = estimator_.inverse_transform(estimator_.transform(imputed))[X_nan]
return X
def _run_TR_from_scan_onsets(self, n_T, scan_onsets=None):
if scan_onsets is None:
# assume that all data are acquired within the same scan.
n_run = 1
run_TRs = np.array([n_T], dtype=int)
else:
# Each value in the scan_onsets tells the index at which
# a new scan starts. For example, if n_T = 500, and
# scan_onsets = [0,100,200,400], this means that the time points
# of 0-99 are from the first scan, 100-199 are from the second,
# 200-399 are from the third and 400-499 are from the fourth
run_TRs = np.int32(np.diff(np.append(scan_onsets, n_T)))
run_TRs = np.delete(run_TRs, np.where(run_TRs == 0))
n_run = run_TRs.size
# delete run length of 0 in case of duplication in scan_onsets.
logger.info('I infer that the number of volumes'
' in each scan are: {}'.format(run_TRs))
return run_TRs, n_run
def chooseErrorData(self, game, lesson=None):
'''
Choose saved error function data by lesson and game name in
history database.
'''
self.history.setGame(game)
self.load()
if lesson is not None:
self.error_data_training = np.split(self.data[0,:],
np.argwhere(self.data[0,:] == -1))[lesson][1:]
self.error_data_test = np.split(self.data[1,:],
np.argwhere(self.data[1,:] == -1))[lesson][1:]
else:
self.error_data_training = np.delete(self.data[0,:],
np.argwhere(self.data[0,:]==-1))
self.error_data_test = np.delete(self.data[1,:],
np.argwhere(self.data[1,:]==-1))
# ------------------- for test and show reasons only ----------------------
def append_neg_and_retrain(self, feat=None, force=False):
if feat is not None:
num = feat.shape[0]
self.neg = np.vstack((self.neg, feat))
self.num_neg_added += num
if self.num_neg_added > self.retrain_limit or force:
self.num_neg_added = 0
new_w_b, pos_scores, neg_scores = self.train()
# scores = np.dot(self.neg, new_w_b[0].T) + new_w_b[1]
# easy_inds = np.where(neg_scores < self.evict_thresh)[0]
not_easy_inds = np.where(neg_scores >= self.evict_thresh)[0]
if len(not_easy_inds) > 0:
self.neg = self.neg[not_easy_inds, :]
# self.neg = np.delete(self.neg, easy_inds)
print(' Pruning easy negatives')
print(' Cache holds {} pos examples and {} neg examples'.
format(self.pos.shape[0], self.neg.shape[0]))
print(' {} pos support vectors'.format((pos_scores <= 1).sum()))
print(' {} neg support vectors'.format((neg_scores >= -1).sum()))
return new_w_b
else:
return None
def add_state(self, state):
if state is None:
self.queue = None
return
state = np.asarray(state)
axis = len(state.shape) # extra dimension for observation
observation = np.reshape(state, state.shape + (1,))
if self.queue is None:
self.queue = np.repeat(observation, self.stacked_num, axis=axis)
else:
# remove oldest observation from the beginning of the observation queue
self.queue = np.delete(self.queue, 0, axis=axis)
# append latest observation to the end of the observation queue
self.queue = np.append(self.queue, observation, axis=axis)
def margins(doc_scores):
margin_win = np.zeros_like(doc_scores)
margin_lose = np.zeros_like(doc_scores)
for j in range(doc_scores.shape[1]):
my_scores = doc_scores[:, j]
others = np.delete(doc_scores, j, axis=1)
if FROM == 'second':
margin_win[:, j] = np.maximum(my_scores - others.max(axis=1), 0)
margin_lose[:, j] = np.maximum(others.min(axis=1) - my_scores, 0)
if FROM == 'other':
margin_win[:, j] = np.maximum(my_scores - others.min(axis=1), 0)
margin_lose[:, j] = np.maximum(others.max(axis=1) - my_scores, 0)
elif FROM == 'median':
margin_win[:, j] = np.maximum(my_scores - np.median(others,
axis=1), 0)
margin_lose[:, j] = np.maximum(np.median(others, axis=1) -
my_scores, 0)
return margin_win, margin_lose
