def fit(
self,
doc, # type: str
predict_proba, # type: Callable[[Any], Any]
):
# type: (...) -> TextExplainer
"""
Explain ``predict_proba`` probabilistic classification function
for the ``doc`` example. This method fits a local classification
pipeline following LIME approach.
To get the explanation use :meth:`show_prediction`,
:meth:`show_weights`, :meth:`explain_prediction` or
:meth:`explain_weights`.
Parameters
----------
doc : str
Text to explain
predict_proba : callable
Black-box classification pipeline. ``predict_proba``
should be a function which takes a list of strings (documents)
and return a matrix of shape ``(n_samples, n_classes)`` with
probability values - a row per document and a column per output
label.
"""
self.doc_ = doc
if self.position_dependent:
samples, sims, mask, text = self.sampler.sample_near_with_mask(
doc=doc, n_samples=self.n_samples)
self.vec_ = SingleDocumentVectorizer(
token_pattern=self.token_pattern).fit([doc])
X = ~mask
else:
self.vec_ = clone(self.vec).fit([doc])
samples, sims = self.sampler.sample_near(doc=doc,
n_samples=self.n_samples)
X = self.vec_.transform(samples)
if self.rbf_sigma is not None:
sims = rbf(1 - sims, sigma=self.rbf_sigma)
self.samples_ = samples
self.similarity_ = sims
self.X_ = X
self.y_proba_ = predict_proba(samples)
self.clf_ = clone(self.clf)
self.metrics_ = _train_local_classifier(
estimator=self.clf_,
samples=X,
similarity=sims,
y_proba=self.y_proba_,
expand_factor=self.expand_factor,
random_state=self.rng_)
return self
def _similarity(self, doc, samples):
distance = _distances(doc, samples, metric=self.metric)
return rbf(distance, sigma=self.sigma_)
