def _mean_var(frame, weights=None):
"""
Compute the (weighted) mean and variance.
:param frame: Single column H2OFrame
:param weights: optional weights column
:returns: The (weighted) mean and variance
"""
return _colmean(frame), frame.var()
def _mean_var(frame, weights=None):
"""
Compute the (weighted) mean and variance
:param frame: Single column H2OFrame
:param weights: optional weights column
:return: The (weighted) mean and variance
"""
return _colmean(frame), frame.var()
def h2o_mean_squared_error(y_actual, y_predicted, weights=None):
"""
Mean squared error regression loss
:param y_actual: H2OFrame of actual response.
:param y_predicted: H2OFrame of predicted response.
:param weights: (Optional) sample weights
:returns: mean squared error loss (best is 0.0).
"""
ModelBase._check_targets(y_actual, y_predicted)
return _colmean((y_predicted - y_actual)**2)
def h2o_mean_squared_error(y_actual, y_predicted, weights=None):
"""
Mean squared error regression loss
:param y_actual: H2OFrame of actual response.
:param y_predicted: H2OFrame of predicted response.
:param weights: (Optional) sample weights
:return: loss (float) (best is 0.0)
"""
ModelBase._check_targets(y_actual, y_predicted)
return _colmean((y_predicted - y_actual) ** 2)
def h2o_mean_absolute_error(y_actual, y_predicted, weights=None):
"""
Mean absolute error regression loss.
:param y_actual: H2OFrame of actual response.
:param y_predicted: H2OFrame of predicted response.
:param weights: (Optional) sample weights
:return: loss (float) (best is 0.0)
"""
ModelBase._check_targets(y_actual, y_predicted)
return _colmean((y_predicted - y_actual).abs())
def h2o_r2_score(y_actual, y_predicted, weights=1.):
"""
R-squared (coefficient of determination) regression score function
:param y_actual: H2OFrame of actual response.
:param y_predicted: H2OFrame of predicted response.
:param weights: (Optional) sample weights
:returns: R-squared (best is 1.0, lower is worse).
"""
ModelBase._check_targets(y_actual, y_predicted)
numerator = (weights * (y_actual - y_predicted)**2).sum()
denominator = (weights * (y_actual - _colmean(y_actual))**2).sum()
if denominator == 0.0:
return 1. if numerator == 0. else 0. # 0/0 => 1, else 0
return 1 - numerator / denominator
def h2o_r2_score(y_actual, y_predicted, weights=1.):
"""
R^2 (coefficient of determination) regression score function
:param y_actual: H2OFrame of actual response.
:param y_predicted: H2OFrame of predicted response.
:param weights: (Optional) sample weights
:return: R^2 (float) (best is 1.0, lower is worse)
"""
ModelBase._check_targets(y_actual, y_predicted)
numerator = (weights * (y_actual - y_predicted) ** 2).sum()
denominator = (weights * (y_actual - _colmean(y_actual)) ** 2).sum()
if denominator == 0.0:
return 1. if numerator == 0. else 0. # 0/0 => 1, else 0
return 1 - numerator / denominator
