Fix rbf kernel calculation

Author: jaime-cespedes-sisniega

frouros/utils/kernels.py

@@ -0,0 +1,21 @@
+"""Kernels module."""
+
+import numpy as np  # type: ignore
+from scipy.spatial.distance import cdist  # type: ignore
+
+
+def rbf_kernel(
+    X: np.ndarray, Y: np.ndarray, sigma: float = 1.0  # noqa: N803
+) -> np.ndarray:
+    """Radial basis function kernel between X and Y matrices.
+
+    :param X: X matrix
+    :type X: numpy.ndarray
+    :param Y: Y matrix
+    :type Y: numpy.ndarray
+    :param sigma: sigma value (equivalent to gamma = 1 / (2 * sigma**2))
+    :type sigma: float
+    :return: Radial basis kernel matrix
+    :rtype: numpy.ndarray
+    """
+    return np.exp(-cdist(X, Y, "sqeuclidean") / (2 * sigma**2))