Support vector lengthscales for RBF and Matern kernels (#1819)

* Support vector lengthscales for RBF and Matern kernels * Use broadcast_shapes to align params * Remove union shorthand * Update test/contrib/hsgp/test_approximation.py Co-authored-by: Juan Orduz <[email protected]> * Run make format * Remove union in isinstance check --------- Co-authored-by: Sam Anklesaria <[email protected]> Co-authored-by: Juan Orduz <[email protected]>
pyro-ppl · Jun 25, 2024 · 2984b9b · 2984b9b
1 parent 209dad9
commit 2984b9b
Show file tree

Hide file tree

Showing 4 changed files with 120 additions and 32 deletions.
diff --git a/numpyro/contrib/hsgp/laplacian.py b/numpyro/contrib/hsgp/laplacian.py
@@ -7,14 +7,13 @@
 
 from __future__ import annotations
 
-from typing import Union, get_args
+from typing import get_args
 
 from jaxlib.xla_extension import ArrayImpl
-import numpy as np
 
 import jax.numpy as jnp
 
-ARRAY_TYPE = Union[ArrayImpl, np.ndarray]
+from numpyro.contrib.hsgp.util import ARRAY_TYPE
 
 
 def eigenindices(m: list[int] | int, dim: int) -> ArrayImpl:

diff --git a/numpyro/contrib/hsgp/spectral_densities.py b/numpyro/contrib/hsgp/spectral_densities.py
@@ -16,8 +16,12 @@
 from numpyro.contrib.hsgp.laplacian import sqrt_eigenvalues
 
 
+def align_param(dim, param):
+    return jnp.broadcast_to(param, jnp.broadcast_shapes(jnp.shape(param), (dim,)))
+
+
 def spectral_density_squared_exponential(
-    dim: int, w: ArrayImpl, alpha: float, length: float
+    dim: int, w: ArrayImpl, alpha: float, length: float | ArrayImpl
 ) -> float:
     """
     Spectral density of the squared exponential kernel.
@@ -44,13 +48,14 @@ def spectral_density_squared_exponential(
     :return: spectral density value
     :rtype: float
     """
-    c = alpha * (jnp.sqrt(2 * jnp.pi) * length) ** dim
-    e = jnp.exp(-0.5 * (length**2) * jnp.dot(w, w))
+    length = align_param(dim, length)
+    c = alpha * jnp.prod(jnp.sqrt(2 * jnp.pi) * length, axis=-1)
+    e = jnp.exp(-0.5 * jnp.sum(w**2 * length**2, axis=-1))
     return c * e
 
 
 def spectral_density_matern(
-    dim: int, nu: float, w: ArrayImpl, alpha: float, length: float
+    dim: int, nu: float, w: ArrayImpl, alpha: float, length: float | ArrayImpl
 ) -> float:
     """
     Spectral density of the Matérn kernel.
@@ -79,22 +84,23 @@ def spectral_density_matern(
     :return: spectral density value
     :rtype: float
     """  # noqa: E501
+    length = align_param(dim, length)
     c1 = (
         alpha
         * (2 ** (dim))
         * (jnp.pi ** (dim / 2))
         * ((2 * nu) ** nu)
         * special.gamma(nu + dim / 2)
     )
-    c2 = (2 * nu / (length**2) + jnp.dot(w, w)) ** (-nu - dim / 2)
-    c3 = special.gamma(nu) * length ** (2 * nu)
+    s = jnp.sum(length**2 * w**2, axis=-1)
+    c2 = jnp.prod(length, axis=-1) * (2 * nu + s) ** (-nu - dim / 2)
+    c3 = special.gamma(nu)
     return c1 * c2 / c3
 
 
-# TODO support length-D kernel hyperparameters
 def diag_spectral_density_squared_exponential(
     alpha: float,
-    length: float,
+    length: float | list[float],
     ell: float | int | list[float | int],
     m: int | list[int],
     dim: int,

diff --git a/numpyro/contrib/hsgp/util.py b/numpyro/contrib/hsgp/util.py
@@ -0,0 +1,10 @@
+# Copyright Contributors to the Pyro project.
+# SPDX-License-Identifier: Apache-2.0
+
+from typing import Union
+
+import numpy as np
+
+import jax
+
+ARRAY_TYPE = Union[jax.Array, np.ndarray]  # jax.Array covers tracers
diff --git a/test/contrib/hsgp/test_approximation.py b/test/contrib/hsgp/test_approximation.py
@@ -5,7 +5,7 @@
 
 from functools import reduce
 from operator import mul
-from typing import Literal
+from typing import Literal, Union
 
 import numpy as np
 import pytest
@@ -74,23 +74,54 @@ def synthetic_two_dim_data() -> tuple[ArrayImpl, ArrayImpl]:
 
 
 @pytest.mark.parametrize(
-    argnames="x1, x2, length, ell",
+    argnames="x1, x2, length, ell, xfail",
     argvalues=[
-        (np.array([[1.0]]), np.array([[0.0]]), np.array([1.0]), 5.0),
+        (np.array([[1.0]]), np.array([[0.0]]), 1.0, 5.0, False),
         (
             np.array([[1.5, 1.25]]),
             np.array([[0.0, 0.0]]),
-            np.array([1.0]),
+            1.0,
+            5.0,
+            False,
+        ),
+        (
+            np.array([[1.5, 1.25]]),
+            np.array([[0.0, 0.0]]),
+            np.array([1.0, 0.5]),
             5.0,
+            False,
+        ),
+        (
+            np.array([[1.5, 1.25]]),
+            np.array([[0.0, 0.0]]),
+            np.array(
+                [[1.0, 0.5], [0.5, 1.0]]
+            ),  # different length scale for each point/dimension
+            5.0,
+            False,
+        ),
+        (
+            np.array([[1.5, 1.25, 1.0]]),
+            np.array([[0.0, 0.0, 0.0]]),
+            np.array([[1.0, 0.5], [0.5, 1.0]]),  # invalid length scale
+            5.0,
+            True,
         ),
     ],
     ids=[
-        "1d",
-        "2d,1d-length",
+        "1d,scalar-length",
+        "2d,scalar-length",
+        "2d,vector-length",
+        "2d,matrix-length",
+        "2d,invalid-length",
     ],
 )
 def test_kernel_approx_squared_exponential(
-    x1: ArrayImpl, x2: ArrayImpl, length: ArrayImpl, ell: float
+    x1: ArrayImpl,
+    x2: ArrayImpl,
+    length: Union[float, ArrayImpl],
+    ell: float,
+    xfail: bool,
 ):
     """ensure that the approximation of the squared exponential kernel is accurate,
     matching the exact kernel implementation from sklearn.
@@ -100,13 +131,26 @@ def test_kernel_approx_squared_exponential(
     assert x1.shape == x2.shape
     m = 100  # large enough to ensure the approximation is accurate
     dim = x1.shape[-1]
+    if xfail:
+        with pytest.raises(ValueError):
+            diag_spectral_density_squared_exponential(1.0, length, ell, m, dim)
+        return
     spd = diag_spectral_density_squared_exponential(1.0, length, ell, m, dim)
 
     eig_f1 = eigenfunctions(x1, ell=ell, m=m)
     eig_f2 = eigenfunctions(x2, ell=ell, m=m)
     approx = (eig_f1 * eig_f2) @ spd
-    exact = RBF(length)(x1, x2)
-    assert jnp.isclose(approx, exact, rtol=1e-3)
+
+    def _exact_rbf(length):
+        return RBF(length)(x1, x2).squeeze(axis=-1)
+
+    if isinstance(length, int) | isinstance(length, float):
+        exact = _exact_rbf(length)
+    elif length.ndim == 1:
+        exact = _exact_rbf(length)
+    else:
+        exact = np.apply_along_axis(_exact_rbf, axis=0, arr=length)
+    assert jnp.isclose(approx, exact, rtol=1e-3).all()
 
 
 @pytest.mark.parametrize(
@@ -118,14 +162,32 @@ def test_kernel_approx_squared_exponential(
             np.array([[1.5, 1.25]]),
             np.array([[0.0, 0.0]]),
             3 / 2,
-            np.array([1.0]),
+            np.array([0.25, 0.5]),
+            5.0,
+        ),
+        (
+            np.array([[1.5, 1.25]]),
+            np.array([[0.0, 0.0]]),
+            5 / 2,
+            np.array([0.25, 0.5]),
+            5.0,
+        ),
+        (
+            np.array([[1.5, 1.25]]),
+            np.array([[0.0, 0.0]]),
+            3 / 2,
+            np.array(
+                [[1.0, 0.5], [0.5, 1.0]]
+            ),  # different length scale for each point/dimension
             5.0,
         ),
         (
             np.array([[1.5, 1.25]]),
             np.array([[0.0, 0.0]]),
             5 / 2,
-            np.array([1.0]),
+            np.array(
+                [[1.0, 0.5], [0.5, 1.0]]
+            ),  # different length scale for each point/dimension
             5.0,
         ),
     ],
@@ -134,6 +196,8 @@ def test_kernel_approx_squared_exponential(
         "1d,nu=5/2",
         "2d,nu=3/2,1d-length",
         "2d,nu=5/2,1d-length",
+        "2d,nu=3/2,2d-length",
+        "2d,nu=5/2,2d-length",
     ],
 )
 def test_kernel_approx_squared_matern(
@@ -154,8 +218,17 @@ def test_kernel_approx_squared_matern(
     eig_f1 = eigenfunctions(x1, ell=ell, m=m)
     eig_f2 = eigenfunctions(x2, ell=ell, m=m)
     approx = (eig_f1 * eig_f2) @ spd
-    exact = Matern(length_scale=length, nu=nu)(x1, x2)
-    assert jnp.isclose(approx, exact, rtol=1e-3)
+
+    def _exact_matern(length):
+        return Matern(length_scale=length, nu=nu)(x1, x2).squeeze(axis=-1)
+
+    if isinstance(length, float) | isinstance(length, int):
+        exact = _exact_matern(length)
+    elif length.ndim == 1:
+        exact = _exact_matern(length)
+    else:
+        exact = np.apply_along_axis(_exact_matern, axis=0, arr=length)
+    assert jnp.isclose(approx, exact, rtol=1e-3).all()
 
 
 @pytest.mark.parametrize(
@@ -211,8 +284,8 @@ def test_approximation_squared_exponential(
     x: ArrayImpl,
     alpha: float,
     length: float,
-    ell: int | float | list[int | float],
-    m: int | list[int],
+    ell: Union[int, float, list[Union[int, float]]],
+    m: Union[int, list[int]],
     non_centered: bool,
 ):
     def model(x, alpha, length, ell, m, non_centered):
@@ -263,8 +336,8 @@ def test_approximation_matern(
     nu: float,
     alpha: float,
     length: float,
-    ell: int | float | list[int | float],
-    m: int | list[int],
+    ell: Union[int, float, list[Union[int, float]]],
+    m: Union[int, list[int]],
     non_centered: bool,
 ):
     def model(x, nu, alpha, length, ell, m, non_centered):
@@ -306,8 +379,8 @@ def model(x, nu, alpha, length, ell, m, non_centered):
 def test_squared_exponential_gp_model(
     synthetic_one_dim_data,
     synthetic_two_dim_data,
-    ell: float | int | list[float | int],
-    m: int | list[int],
+    ell: Union[float, int, list[Union[float, int]]],
+    m: Union[int, list[int]],
     non_centered: bool,
     num_dim: Literal[1, 2],
 ):
@@ -364,8 +437,8 @@ def test_matern_gp_model(
     synthetic_one_dim_data,
     synthetic_two_dim_data,
     nu: float,
-    ell: int | float | list[float | int],
-    m: int | list[int],
+    ell: Union[int, float, list[Union[float, int]]],
+    m: Union[int, list[int]],
     non_centered: bool,
     num_dim: Literal[1, 2],
 ):