Add laplacian kernel

1 files changed, 55 insertions, 0 deletions

diff --git a/ml_exp/kernels.py b/ml_exp/kernels.py
index abc71f7af..488232392 100644
--- a/ml_exp/kernels.py
+++ b/ml_exp/kernels.py
@@ -83,3 +83,58 @@ def gaussian_kernel(X1,
                 norm = np.linalg.norm(X2 - X1[i], axis=-1)
             K[i, :] = np.exp(i_sigma * np.square(norm))
     return K
+
+
+def laplacian_kernel(X1,
+                     X2,
+                     sigma,
+                     use_tf=True):
+    """
+    Calculates the Laplacian Kernel.
+    X1: first representations.
+    X2: second representations.
+    sigma: kernel width.
+    use_tf: if tensorflow should be used.
+    """
+    # If tf is to be used but couldn't be imported, don't try to use it.
+    if use_tf and not TF_AV:
+        use_tf = False
+
+    X1_size = X1.shape[0]
+    X2_size = X2.shape[0]
+    i_sigma = -0.5 / sigma
+
+    if use_tf:
+        if tf.config.experimental.list_physical_devices('GPU'):
+            with tf.device('GPU:0'):
+                X1 = tf.convert_to_tensor(X1)
+                X2 = tf.convert_to_tensor(X2)
+                X2r = tf.rank(X2)
+
+                def cond(i, _):
+                    return tf.less(i, X1_size)
+
+                def body(i, K):
+                    if X2r == 3:
+                        norm = tf.norm(X2 - X1[i], axis=(1, 2))
+                    else:
+                        norm = tf.norm(X2 - X1[i], axis=-1)
+
+                    return (i + 1,
+                            K.write(i, tf.exp(i_sigma * norm)))
+
+                K = tf.TensorArray(dtype=tf.float64,
+                                   size=X1_size)
+                i_state = (0, K)
+                n, K = tf.while_loop(cond, body, i_state)
+                K = K.stack()
+    else:
+        K = np.zeros((X1_size, X2_size), dtype=np.float64)
+        # Faster way of calculating the kernel (no numba support).
+        for i in range(X1_size):
+            if X2.ndim == 3:
+                norm = np.linalg.norm(X2 - X1[i], axis=(1, 2))
+            else:
+                norm = np.linalg.norm(X2 - X1[i], axis=-1)
+            K[i, :] = np.exp(i_sigma * norm)
+    return K