Anchor Tabular beam size not used

[smwrig]

Hi, it seems the beam_size parameter of AnchorTabularExplainer.explain_instance doesn't work as intended. It doesn't get passed to the anchor_beam method, which defaults to 1 and limits the search.

anchor/anchor/anchor_tabular.py

Lines 269 to 281 in b1f5e6c

	def explain_instance(self, data_row, classifier_fn, threshold=0.95,
	delta=0.1, tau=0.15, batch_size=100,
	max_anchor_size=None,
	desired_label=None,
	beam_size=4, **kwargs):
	# It's possible to pass in max_anchor_size
	sample_fn, mapping = self.get_sample_fn(
	data_row, classifier_fn, desired_label=desired_label)
	# return sample_fn, mapping
	exp = anchor_base.AnchorBaseBeam.anchor_beam(
	sample_fn, delta=delta, epsilon=tau, batch_size=batch_size,
	desired_confidence=threshold, max_anchor_size=max_anchor_size,
	**kwargs)

anchor/anchor/anchor_base.py

Lines 255 to 259 in b1f5e6c

	def anchor_beam(sample_fn, delta=0.05, epsilon=0.1, batch_size=10,
	min_shared_samples=0, desired_confidence=1, beam_size=1,
	verbose=False, epsilon_stop=0.05, min_samples_start=0,
	max_anchor_size=None, verbose_every=1,
	stop_on_first=False, coverage_samples=10000):