def fit(self, X, y=None, **fit_params):
"""
Parameters
----------
X: sparse matrix of shape (n_docs, n_words)
The data matrix to used to find the low-rank effects
y: Ignored
fit_params:
optional model params
Returns
-------
self
"""
if self.model_type == "pLSA":
self.model_ = enstop.PLSA(n_components=self.n_components,
**fit_params).fit(X)
elif self.model_type == "EnsTop":
self.model_ = enstop.EnsembleTopics(n_components=self.n_components,
**fit_params).fit(X)
else:
raise ValueError("model_type is not supported")
return self
def fit(self, X, y=None, **fit_params):
"""
Parameters
----------
X: sparse matrix of shape (n_docs, n_words)
The data matrix that get's binarized that the model is attempting to fit to.
y: Ignored
fit_params:
optional model params
Returns
-------
self
"""
binary_indicator_matrix = (X != 0).astype(np.float32)
if self.model_type == "pLSA":
self.model_ = enstop.PLSA(n_components=self.n_components, **fit_params).fit(
binary_indicator_matrix
)
elif self.model_type == "EnsTop":
self.model_ = enstop.EnsembleTopics(
n_components=self.n_components, **fit_params
).fit(binary_indicator_matrix)
else:
raise ValueError("model_type is not supported")
return self
def fit(self, X, y=None, **fit_params):
"""
Parameters
----------
X: sparse matrix of shape (n_docs, n_words)
The data matrix (that potentially get's binarized) that the model is attempting to fit to.
y: Ignored
fit_params:
optional model params
Returns
-------
self
"""
if callable(self.information_function):
self._information_function = self.information_function
elif self.information_function in _INFORMATION_FUNCTIONS:
self._information_function = _INFORMATION_FUNCTIONS[
self.information_function]
else:
raise ValueError(
f"Unrecognized kernel_function; should be callable or one of {_INFORMATION_FUNCTIONS.keys()}"
)
if self.information_function in ["idf", "average idf"]:
self.binarize_matrix = True
elif self.information_function in ["column KL", "Bernoulli KL"]:
self.binarize_matrix = False
if self.binarize_matrix:
binary_indicator_matrix = (X != 0).astype(np.float32)
if self.model_type == "pLSA":
self.model_ = enstop.PLSA(
n_components=self.n_components,
**fit_params).fit(binary_indicator_matrix)
elif self.model_type == "EnsTop":
self.model_ = enstop.EnsembleTopics(
n_components=self.n_components,
**fit_params).fit(binary_indicator_matrix)
else:
raise ValueError("model_type is not supported")
self.token_counts_ = np.array(
binary_indicator_matrix.sum(axis=1)).T[0]
else:
if self.model_type == "pLSA":
self.model_ = enstop.PLSA(n_components=self.n_components,
**fit_params).fit(
X.astype(np.float32))
elif self.model_type == "EnsTop":
self.model_ = enstop.EnsembleTopics(
n_components=self.n_components,
**fit_params).fit(X.astype(np.float32))
else:
raise ValueError("model_type is not supported")
return self