def compute_scores(self,
X: np.ndarray,
y: np.ndarray,
name: str = "") -> pd.Series:
scores = pd.Series()
pred = self.predict(X)
scores.loc["RMSE{}".format(name)] = rmse(pred, y)
return scores
def compute_scores(self, X: np.ndarray, y: np.ndarray,
name: str = "") -> pd.Series:
scores = pd.Series()
pred = self.predict(X)
X_array = self.extract_subdatasets(X)
len_sequences = np.sum(((X_array[1] > -10)*1)[:, :, 1], axis=1)
scores.loc["RMSE{}".format(name)] = rmse(pred[:, 1], y)
scores.loc["conf_interval_prop_{}"
.format(name)] = conf_interval_prop(y, pred)
scores.loc["conf_interval_size_{}"
.format(name)] = conf_interval_size(pred)
idx_inf = len_sequences < 3
scores.loc["rmse_inf{}".format(name)] = rmse(pred[idx_inf, 1],
y[idx_inf])
idx_sup = len_sequences >= 3
scores.loc["rmse_sup{}".format(name)] = rmse(pred[idx_sup, 1],
y[idx_sup])
return scores
def score(self, X: np.ndarray, y: np.ndarray) -> float:
"""
Computes rmse between target y and predictions for input X after
extracting different subdatasets (image, scalar, and constant).
Args:
X (np.ndarray): Source array, containing all the type of
inputs, concatenated in one dataframe.
y (np.ndarray): Target
Returns:
float: rmse of predictions
"""
X = self.extract_subdatasets(X)
pred = self.model.predict(X)
return rmse(pred, y)