def fit(self, train_set, val_set=None):
"""Fit the model to observations.
Parameters
----------
train_set: :obj:`cornac.data.Dataset`, required
User-Item preference data as well as additional modalities.
val_set: :obj:`cornac.data.Dataset`, optional, default: None
User-Item preference data for model selection purposes (e.g., early stopping).
Returns
-------
self : object
"""
Recommender.fit(self, train_set, val_set)
X = sp.csc_matrix(self.train_set.matrix)
# recover the striplet sparse format from csc sparse matrix X (needed to feed c++)
(rid, cid, val) = sp.find(X)
val = np.array(val, dtype='float32')
rid = np.array(rid, dtype='int32')
cid = np.array(cid, dtype='int32')
tX = np.concatenate((np.concatenate(
([rid], [cid]), axis=0).T, val.reshape((len(val), 1))),
axis=1)
del rid, cid, val
if self.trainable:
map_iid = train_set.iid_list
(rid, cid,
val) = train_set.item_graph.get_train_triplet(map_iid, map_iid)
context_info = np.hstack(
(rid.reshape(-1, 1), cid.reshape(-1, 1), val.reshape(-1, 1)))
if self.variant == 'c2pf':
res = c2pf.c2pf(tX, X.shape[0], X.shape[1], context_info,
X.shape[1], X.shape[1], self.k, self.max_iter,
self.init_params)
elif self.variant == 'tc2pf':
res = c2pf.t_c2pf(tX, X.shape[0], X.shape[1], context_info,
X.shape[1], X.shape[1], self.k,
self.max_iter, self.init_params)
elif self.variant == 'rc2pf':
res = c2pf.r_c2pf(tX, X.shape[0], X.shape[1], context_info,
X.shape[1], X.shape[1], self.k,
self.max_iter, self.init_params)
else:
res = c2pf.c2pf(tX, X.shape[0], X.shape[1], context_info,
X.shape[1], X.shape[1], self.k, self.max_iter,
self.init_params)
self.Theta = sp.csc_matrix(res['Z']).todense()
self.Beta = sp.csc_matrix(res['W']).todense()
self.Xi = sp.csc_matrix(res['Q']).todense()
elif self.verbose:
print('%s is trained already (trainable = False)' % (self.name))
return self
def fit(self, train_set):
"""Fit the model to observations.
Parameters
----------
train_set: object of type TrainSet, required
An object contraining the user-item preference in csr scipy sparse format,\
as well as some useful attributes such as mappings to the original user/item ids.\
Please refer to the class TrainSet in the "data" module for details.
"""
Recommender.fit(self, train_set)
X = sp.csc_matrix(self.train_set.matrix)
# recover the striplet sparse format from csc sparse matrix X (needed to feed c++)
(rid, cid, val) = sp.find(X)
val = np.array(val, dtype='float32')
rid = np.array(rid, dtype='int32')
cid = np.array(cid, dtype='int32')
tX = np.concatenate((np.concatenate(
([rid], [cid]), axis=0).T, val.reshape((len(val), 1))),
axis=1)
del rid, cid, val
if self.trainable:
map_iid = train_set.iid_list
(rid, cid,
val) = train_set.item_graph.get_train_triplet(map_iid, map_iid)
context_info = np.hstack(
(rid.reshape(-1, 1), cid.reshape(-1, 1), val.reshape(-1, 1)))
if self.variant == 'c2pf':
res = c2pf.c2pf(tX, X.shape[0], X.shape[1], context_info,
X.shape[1], X.shape[1], self.k, self.max_iter,
self.init_params)
elif self.variant == 'tc2pf':
res = c2pf.t_c2pf(tX, X.shape[0], X.shape[1], context_info,
X.shape[1], X.shape[1], self.k,
self.max_iter, self.init_params)
elif self.variant == 'rc2pf':
res = c2pf.r_c2pf(tX, X.shape[0], X.shape[1], context_info,
X.shape[1], X.shape[1], self.k,
self.max_iter, self.init_params)
else:
res = c2pf.c2pf(tX, X.shape[0], X.shape[1], context_info,
X.shape[1], X.shape[1], self.k, self.max_iter,
self.init_params)
self.Theta = sp.csc_matrix(res['Z']).todense()
self.Beta = sp.csc_matrix(res['W']).todense()
self.Xi = sp.csc_matrix(res['Q']).todense()
elif self.verbose:
print('%s is trained already (trainable = False)' % (self.name))
def fit(self, train_set, val_set=None):
"""Fit the model to observations.
Parameters
----------
train_set: :obj:`cornac.data.Dataset`, required
User-Item preference data as well as additional modalities.
val_set: :obj:`cornac.data.Dataset`, optional, default: None
User-Item preference data for model selection purposes (e.g., early stopping).
Returns
-------
self : object
"""
Recommender.fit(self, train_set, val_set)
X = sp.csc_matrix(self.train_set.matrix)
# recover the striplet sparse format from csc sparse matrix X (needed to feed c++)
(rid, cid, val) = sp.find(X)
val = np.array(val, dtype="float32")
rid = np.array(rid, dtype="int32")
cid = np.array(cid, dtype="int32")
tX = np.concatenate(
(np.concatenate(
([rid], [cid]), axis=0).T, val.reshape((len(val), 1))),
axis=1,
)
del rid, cid, val
if self.trainable:
# use pre-trained params if exists, otherwise from constructor
init_params = {
"G_s": self.Gs,
"G_r": self.Gr,
"L_s": self.Ls,
"L_r": self.Lr,
"L2_s": self.L2s,
"L2_r": self.L2r,
"L3_s": self.L3s,
"L3_r": self.L3r,
}
map_iid = train_set.item_indices
(rid, cid,
val) = train_set.item_graph.get_train_triplet(map_iid, map_iid)
context_info = np.hstack(
(rid.reshape(-1, 1), cid.reshape(-1, 1), val.reshape(-1, 1)))
if self.variant == "c2pf":
res = c2pf.c2pf(
tX,
X.shape[0],
X.shape[1],
context_info,
X.shape[1],
X.shape[1],
self.k,
self.max_iter,
init_params,
)
elif self.variant == "tc2pf":
res = c2pf.t_c2pf(
tX,
X.shape[0],
X.shape[1],
context_info,
X.shape[1],
X.shape[1],
self.k,
self.max_iter,
init_params,
)
elif self.variant == "rc2pf":
res = c2pf.r_c2pf(
tX,
X.shape[0],
X.shape[1],
context_info,
X.shape[1],
X.shape[1],
self.k,
self.max_iter,
init_params,
)
else:
res = c2pf.c2pf(
tX,
X.shape[0],
X.shape[1],
context_info,
X.shape[1],
X.shape[1],
self.k,
self.max_iter,
init_params,
)
self.Theta = sp.csc_matrix(res["Z"]).todense()
self.Beta = sp.csc_matrix(res["W"]).todense()
self.Xi = sp.csc_matrix(res["Q"]).todense()
# overwrite init_params for future fine-tuning
self.Gs = np.asarray(res["G_s"])
self.Gr = np.asarray(res["G_r"])
self.Ls = np.asarray(res["L_s"])
self.Lr = np.asarray(res["L_r"])
self.L2s = np.asarray(res["L2_s"])
self.L2r = np.asarray(res["L2_r"])
self.L3s = np.asarray(res["L3_s"])
self.L3r = np.asarray(res["L3_r"])
elif self.verbose:
print("%s is trained already (trainable = False)" % (self.name))
return self