gibbs function added

gpytorch/kernels/gibbs_kernel.py

@@ -0,0 +1,32 @@
+#!/usr/bin/env python3
+
+
+import torch
+
+from .kernel import Kernel
+
+
+class GibbsKernel(Kernel):
+    r"""Computes a covariance matrix based on the gibbs kernel
+    between inputs :math:`\mathbf{x_1}` and :math:`\mathbf{x_2}`:
+    """
+
+    has_lengthscale = False
+
+    def __init__(
+        self, lengthscale1, lengthscale2, ard_num_dims: int = 1, batch_shape: torch.Size = torch.Size([]), **kwargs
+    ):
+        super().__init__(ard_num_dims=ard_num_dims, batch_shape=batch_shape, **kwargs)
+        self.lengthscale1 = lengthscale1
+        self.lengthscale2 = lengthscale2
+
+    def forward(self, x1, x2, diag=False, **params):
+        x1_ = x1
+        x2_ = x2
+        diff = (x1_.unsqueeze(-2) - x2_.unsqueeze(-3)).pow(2)
+        square_term = (self.lengthscale1).pow(2).unsqueeze(-2) + (self.lengthscale2).pow(2).unsqueeze(-3)
+        prod_term = 2 * (self.lengthscale1) * (self.lengthscale2)
+        res = (prod_term / square_term).pow(0.5).prod(dim=-1) * ((-(diff / square_term).sum(dim=-1)).exp_())
+        if diag:
+            res = res.squeeze(0)
+        return res