fdars.conformal

Conformal prediction and inference methods for functional data: regression intervals, nonparametric intervals, and classification prediction sets.

Functions

Function	Description
conformal_fregre_lm	Conformal prediction intervals for linear regression
conformal_fregre_np	Conformal prediction intervals for nonparametric regression
conformal_classif	Conformal classification prediction sets

---

conformal_fregre_lm

fdars.conformal_fregre_lm(data, response, test_data, ncomp=3,
                          cal_fraction=0.25, alpha=0.1, seed=42)

Split conformal prediction intervals for scalar-on-function linear regression.

Parameter	Type	Default	Description
data	ndarray (n, m)		Training functional predictors
response	ndarray (n,)		Scalar response
test_data	ndarray (n_test, m)		Test functional predictors
ncomp	int	3	Number of FPC components
cal_fraction	float	0.25	Fraction of training data used for calibration
alpha	float	0.1	Miscoverage level (1 - confidence)
seed	int	42	Random seed for train/calibration split

Returns	Type	Description
result	dict	Keys: lower (n_test,), upper (n_test,), predictions (n_test,), coverage

result = fdars.conformal_fregre_lm(X_train, y_train, X_test, ncomp=5, alpha=0.1)
for i in range(len(result["predictions"])):
    print(f"  [{result['lower'][i]:.2f}, {result['upper'][i]:.2f}]")

---

conformal_fregre_np

fdars.conformal_fregre_np(data, response, test_data, argvals,
                          cal_fraction=0.25, alpha=0.1,
                          h_func=1.0, h_scalar=1.0, seed=42)

Split conformal prediction intervals for nonparametric kernel regression.

Parameter	Type	Default	Description
data	ndarray (n, m)		Training functional predictors
response	ndarray (n,)		Scalar response
test_data	ndarray (n_test, m)		Test functional predictors
argvals	ndarray (m,)		Evaluation points
cal_fraction	float	0.25	Calibration fraction
alpha	float	0.1	Miscoverage level
h_func	float	1.0	Functional bandwidth
h_scalar	float	1.0	Scalar bandwidth
seed	int	42	Random seed

Returns	Type	Description
result	dict	Keys: lower (n_test,), upper (n_test,), predictions (n_test,), coverage

t = np.linspace(0, 1, 100)
result = fdars.conformal_fregre_np(X_train, y_train, X_test, t, alpha=0.1)

---

conformal_classif

fdars.conformal_classif(data, labels, test_data, ncomp=3,
                        classifier="lda", cal_fraction=0.25,
                        alpha=0.1, seed=42)

Conformal classification prediction sets. Returns a set of plausible classes for each test observation with guaranteed coverage.

Parameter	Type	Default	Description
data	ndarray (n, m)		Training functional data
labels	ndarray (n,) of int64		Class labels
test_data	ndarray (n_test, m)		Test functional data
ncomp	int	3	Number of FPC components
classifier	str	"lda"	"lda", "qda", or "knn"
cal_fraction	float	0.25	Calibration fraction
alpha	float	0.1	Miscoverage level
seed	int	42	Random seed

Returns	Type	Description
result	dict	Keys: prediction_sets (list of lists of int), coverage

result = fdars.conformal_classif(X_train, y_train, X_test, classifier="lda")
for i, pset in enumerate(result["prediction_sets"]):
    print(f"  Obs {i}: possible classes = {pset}")