leaf_font_size=12,
p=50,
**kwargs)
df = pd.read_csv('data/cervical_arranged_NORM.csv')
gt_labels = ['Biopsy']
y = df.Biopsy.values.astype(np.int32)
df = df.drop(gt_labels, 1)
X = df.as_matrix().astype(np.float64)
model = MultitaskEmbedding(embedding='zero',
bypass=False,
alpha=0.1,
width=10,
depth=2)
model.fit(X, y)
domains = [np.unique(X[:, i]) for i in range(X.shape[1])]
impact = np.zeros((X.shape[1], 10))
for ft_orig in range(X.shape[1]):
emb = model.get_hidden_representation(X)
for value in domains[ft_orig]:
Xnew = X.copy()
Xnew[:, ft_orig] = value
embnew = model.get_hidden_representation(Xnew)
impact[ft_orig] = np.maximum(impact[ft_orig],
skf = StratifiedKFold(10, shuffle=False, random_state=42)
df = pd.read_csv('data/cervical_arranged_NORM.csv')
all_procedures = set(['Hinselmann', 'Schiller', 'Citology'])
for gt_labels in list(findsubsets(all_procedures)):
gt_labels = list(gt_labels) + ['Biopsy']
print 'Procedures', ' '.join(sorted(list(all_procedures - set(gt_labels))))
y = df.Biopsy.values.astype(np.int32)
X = df.drop(gt_labels, 1).as_matrix().astype(np.float64)
cv = 3
models = [('Sup',
GridSearchCV(
MultitaskEmbedding(alpha=1., embedding='raw'),
param_grid={
'depth': [1, 2, 3],
'width': [10, 20],
},
scoring='average_precision',
cv=cv,
)),
('Semi',
GridSearchCV(
MultitaskEmbedding(alpha=0.01),
param_grid={
'alpha': [0.01, 0.1],
'depth': [1, 2, 3],
'width': [10, 20],
'bypass': [False, True],
ret = metrics.average_precision_score(a, b[:, 1])
if np.isnan(ret):
return -np.inf
return ret
X = pd.read_csv(os.sys.argv[1]).as_matrix()
y = X[:, -1]
X = X[:, : -1]
skf = StratifiedKFold(10, shuffle=False, random_state=42)
cv = 3
models = [('Sup',
GridSearchCV(MultitaskEmbedding(alpha=1.,
embedding='raw'),
param_grid={'depth': [1, 2, 3],
'width': [10, 20],
},
scoring='average_precision',
cv=cv,
)),
('Semi',
GridSearchCV(MultitaskEmbedding(alpha=0.01),
param_grid={'alpha': [0.01, 0.1],
'depth': [1, 2, 3],
'width': [10, 20],
'bypass': [False, True],
},
scoring='average_precision',
cv=cv,
os.sys.setrecursionlimit(100000)
skf = StratifiedKFold(10, shuffle=False, random_state=42)
df = pd.read_csv('data/cervical_arranged_NORM.csv')
all_procedures = set(['Hinselmann', 'Schiller', 'Citology'])
gt_labels = ['Biopsy']
y = df.Biopsy.values.astype(np.int32)
X = df.drop(gt_labels, 1).as_matrix().astype(np.float64)
cv = 3
models = [
('Unsupervised',
GridSearchCV(MultitaskEmbedding(alpha=0.0, bypass=False),
param_grid={
'depth': [1, 2, 3],
'width': [10, 20],
},
scoring='average_precision',
cv=cv,
n_jobs=1)),
('Semi',
GridSearchCV(MultitaskEmbedding(alpha=0.01, bypass=False),
param_grid={
'alpha': [0.01, 0.1],
'depth': [1, 2, 3],
'width': [10, 20],
},
scoring='average_precision',