Source code for caliber.multiclass_classification.ood.da_exp_interpolant
from typing import Any, Optional
import numpy as np
from scipy import stats
from caliber.multiclass_classification.base import AbstractMulticlassClassificationModel
[docs]
class DistanceAwareExponentialInterpolantMulticlassClassificationModel(
AbstractMulticlassClassificationModel
):
def __init__(self, model: Optional[Any] = None, conf_distance: float = 0.99):
super().__init__()
self.model = model
self.conf_distance = conf_distance
self._quantile_distance = None
self._rv = None
[docs]
def fit(self, probs: np.ndarray, distances: np.ndarray, targets: np.ndarray):
if self.model is not None:
self.model.fit(probs, targets)
self._quantile_distance = np.quantile(distances, self.conf_distance)
self._rv = stats.expon(*stats.expon.fit(distances))
[docs]
def predict_proba(self, probs: np.ndarray, distances: np.ndarray) -> np.ndarray:
probs = np.copy(probs)
if self.model is not None:
probs = self.model.predict_proba(probs)
cdf = self._get_cdf(distances)[:, None]
return (1 - cdf) * probs + cdf / probs.shape[1]
[docs]
def predict(self, probs: np.ndarray, distances: np.ndarray) -> np.ndarray:
return np.argmax(self.predict_proba(probs, distances), axis=1)
def _get_cdf(self, distances: np.ndarray) -> np.ndarray:
d = distances - self._quantile_distance
return np.where(d > 0, self._rv.cdf(d), 0.0)