def mock_ks_explainer(request):
"""
Instantiates a KernelShap explainer with a mock predictor.
"""
pred_out_dim, link = request.param
predictor = MockPredictor(out_dim=pred_out_dim, seed=0)
explainer = KernelShap(predictor=predictor, seed=0)
return explainer
def mock_ale_explainer(request):
"""
Instantiates an ALE explainer with a mock predictor.
"""
out_dim, out_type = request.param
predictor = MockPredictor(out_dim=out_dim, out_type=out_type, seed=0)
explainer = ALE(predictor)
return explainer
def patch_transformer(monkeypatch, request):
out_dim, out_type, sz = request.param
transformer = ArgmaxTransformer(None)
monkeypatch.setattr(
transformer, "predictor",
MockPredictor(out_dim,
out_type=out_type,
sz=sz,
key=(out_dim, out_type, sz)))
return transformer
def mock_kernel_shap_explainer(request):
"""
Instantiates a KernelShap explainer with a mock predictor.
"""
pred_out_dim, link, distributed_opts = request.param
predictor = MockPredictor(out_dim=pred_out_dim, seed=0)
explainer = KernelShap(predictor=predictor,
link=link,
distributed_opts=distributed_opts,
seed=0)
return explainer
def mock_tree_shap_explainer(monkeypatch, request):
"""
Instantiates a TreeShap explainer where both the `_explainer` and `model` atttributes
are mocked
Parameters
----------
request:
Tuple containing number of outputs and model_output string (see TreeShap constructor).
"""
n_outs, model_output = request.param
seed = 0
predictor = MockPredictor(out_dim=n_outs, out_type=model_output, seed=seed)
tree_explainer = MockTreeExplainer(predictor, seed=seed)
monkeypatch.setattr(shap, "TreeExplainer", tree_explainer)
explainer = TreeShap(predictor, model_output=model_output)
return explainer
def test_argmax_transformer(monkeypatch, out_dim, out_type):
"""
Test that the conversion of output probabilities to class labels works
for a range of classifier output types.
"""
# setup a transformer with a mock predictor
transformer = ArgmaxTransformer(None)
monkeypatch.setattr(
transformer,
"predictor",
MockPredictor(
out_dim,
out_type=out_type,
)
)
# fake predictor input
X = np.random.random(size=(50, 14))
result = transformer(X)
# argmax transformer should do get rid of the feature dimension
assert len(result.shape) == len(X.shape) - 1