def fit(self, X, y):
"""
X - SPGeneratorData
X[i] = {"id":0,"sp":[[image, ... ],...],"descriptors":[<row of features>]}
"""
sp_list = QSARModel1.getSPList(X)
self.sptype = descr.SPType(n_image_clusters=self.n_image_clusters,
n_prop_clusters=self.n_prop_clusters)
self.sptype.fit(sp_list)
mat = descr.createDescriptors(sp_list, self.sptype)
if "descriptors" in X[0]:
# build features_mat
features_list = []
for el in X:
features_list.append(el["descriptors"])
features_mat = np.array(features_list)
# concatenate mat and features_mat
mat = np.concatenate([mat, features_mat], axis=1)
self.trainer = clone(self.estimator)
self.trainer.fit(mat, y)
return
def fit(self, X, y):
"""
X - SPGeneratorData
X[i] = {"id":0,"data":[mol_points, mol_props],"features":[<row of features>]}
"""
sp_list, segm_list = self.sp_generator_fun(X, self.n_sp_clusters,
self.imsize)
self.sptype = descr.SPType(n_image_clusters=self.n_image_clusters,
n_prop_clusters=self.n_prop_clusters)
self.sptype.fit(sp_list)
mat = descr.createDescriptors(sp_list, self.sptype)
if "features" in X[0]:
# build features_mat
features_list = []
for el in X:
features_list.append(el["features"])
features_mat = np.array(features_list)
# concatenate mat and features_mat
mat = np.concatenate([mat, features_mat], axis=1)
self.trainer = clone(self.estimator)
self.trainer.fit(mat, y)
return
def score(self, X, y, sample_weight=None):
"""
X - SPGeneratorData
X[i] = {"id":0,"sp":[[images, ... ],...],"descriptors":[<row of features>]}
"""
sp_list = QSARModel1.getSPList(X)
mat = descr.createDescriptors(sp_list, self.sptype)
if "descriptors" in X[0]:
# build features_mat
features_list = []
for el in X:
features_list.append(el["descriptors"])
features_mat = np.array(features_list)
# concatenate mat and features_mat
mat = np.concatenate([mat, features_mat], axis=1)
return self.trainer.score(mat, y, sample_weight)
def getMatrix(self, X):
"""
X - SPGeneratorData
return md - matrix
"""
sp_list = QSARModel1.getSPList(X)
self.sptype = descr.SPType(n_image_clusters=self.n_image_clusters,
n_prop_clusters=self.n_prop_clusters)
self.sptype.fit(sp_list)
mat = descr.createDescriptors(sp_list, self.sptype)
if "descriptors" in X[0]:
# build features_mat
features_list = []
for el in X:
features_list.append(el["descriptors"])
features_mat = np.array(features_list)
# concatenate mat and features_mat
mat = np.concatenate([mat, features_mat], axis=1)
return mat
def predict(self, X):
"""
X - SPGeneratorData
X[i] = {"id":0,"data":[mol_points, mol_props],"features":[<row of features>]}
"""
# X[i] <-> sp_list[i]
sp_list, segm_list = self.sp_generator_fun(X, self.n_sp_clusters,
self.imsize)
mat = descr.createDescriptors(sp_list, self.sptype)
if "features" in X[0]:
# build features_mat
features_list = []
for el in X:
features_list.append(el["features"])
features_mat = np.array(features_list)
# concatenate mat and features_mat
mat = np.concatenate([mat, features_mat], axis=1)
return self.trainer.predict(mat)