def predict(data, model, interval=0.95, **kwargs):
def pred(X):
if hasattr(model, 'predict_proba'):
Ey, Vy, ql, qu = model.predict_proba(X, interval, **kwargs)
predres = np.hstack((Ey[:, np.newaxis], Vy[:, np.newaxis],
ql[:, np.newaxis], qu[:, np.newaxis]))
else:
predres = np.reshape(model.predict(X, **kwargs),
newshape=(len(X), 1))
if hasattr(model, 'entropy_reduction'):
MI = model.entropy_reduction(X)
predres = np.hstack((predres, MI[:, np.newaxis]))
if hasattr(model, 'krige_residual'):
kr = model.krige_residual(lon_lat=kwargs['lon_lat'])
predres = np.hstack((predres, kr[:, np.newaxis]))
if hasattr(model, 'ml_prediction'):
ml_pred = model.ml_prediction(X)
predres = np.hstack((predres, ml_pred[:, np.newaxis]))
return predres
result = apply_masked(pred, data)
return result
def test_apply_masked(masked_data):
yt, Xt, ys, Xs = masked_data
yt_masked = np.ma.masked_array(yt, mask=Xt.mask.flatten())
def ident(X):
return X
def target(X):
return yt[~Xt.mask.flatten()]
def fit(X):
assert np.allclose(X, Xt.data[~Xt.mask.flatten()])
return
assert np.ma.all(Xt == apply_masked(ident, Xt))
assert np.ma.all(yt_masked == apply_masked(target, Xt))
assert apply_masked(fit, Xt) is None