class RegularizedGreedyForestEstimator(BaseEstimator):
def __init__(self, max_leaf_nodes=500, l2=0.01, loss='LS',
test_interval=100, verbose=False):
self.max_leaf_nodes = max_leaf_nodes
self.l2 = l2
self.loss = loss
self.test_interval = test_interval
self.verbose = verbose
self.ensemble = None
def _fit(self, X, y, sample_weight=None):
check_weight(sample_weight)
# need to check X is c-contigous!
if self.ensemble is None:
self.ensemble = RGFTreeEnsemble()
builder = RGFBuilder(self.max_leaf_nodes, self.l2, self.loss,
self.test_interval, self.verbose)
builder.build(self.ensemble, X, y, sample_weight)
def fit(self, X, y, sample_weight=None):
self._fit(X, y, sample_weight)
def _validate_X_predict(self, X):
if self.ensemble is None or len(self.ensemble) == 0:
raise NotFittedError("Estimator not fitted, "
"all `fit` before exploiting the model.")
return X
def predict(self, X):
X = self._validate_X_predict(X)
return self.ensemble.predict(X)
def _fit(self, X, y, sample_weight=None):
check_weight(sample_weight)
# need to check X is c-contigous!
if self.ensemble is None:
self.ensemble = RGFTreeEnsemble()
builder = RGFBuilder(self.max_leaf_nodes, self.l2, self.loss,
self.test_interval, self.verbose)
builder.build(self.ensemble, X, y, sample_weight)