Merge pull request #7 from cics-nd/predict-f-samples

Predict f samples

Author: sdatkinson

gptorch/model.py

@@ -495,6 +495,20 @@ def predict_y(self, input_new, diag=True):
         else:
             return self.likelihood.predict_mean_covariance(mean_f, cov_f)
 
+    @input_as_tensor
+    def predict_f_samples(self, input_new, n_samples=1):
+        """
+        Return [n_samp x n_test x d_y] matrix of samples
+        :param input_new:
+        :param n_samples:
+        :return:
+        """
+        mu, sigma = self.predict_f(input_new, diag=False)
+        chol_s = cholesky(sigma)
+        samp = mu + torch.stack([torch.mm(chol_s, Variable(torch.Tensor(r)))
+                              for r in np.random.randn(n_samples, *mu.size())])
+        return samp
+
     @input_as_tensor
     def predict_y_samples(self, input_new, n_samples=1):
         """

test/test_likelihoods.py

@@ -55,7 +55,7 @@ def test_logp(self):
         assert isinstance(logp, Variable)
 
         # Value
-        assert logp.data.numpy() == expected_logp
+        assert logp.data.numpy() == pytest.approx(expected_logp)
 
     def test_predict_mean_variance(self):
         """

test/test_model.py

@@ -12,6 +12,8 @@
 import torch as th
 
 from gptorch.model import Param, Model, GPModel
+from gptorch.kernels import Rbf
+from gptorch.models import GPR
 
 
 class TestParam(object):
@@ -58,4 +60,42 @@ class TestGPModel(object):
     """
     Tests for the GPModel class
     """
-    pass
+    def test_predict_f_samples(self):
+        # TODO mock a GPModel?  Using GPR for the moment.
+        n, dx, dy = 5, 3, 2
+        x, y = np.random.randn(n, dx), np.random.randn(n, dy)
+        kern = Rbf(dx, ARD=True)
+        gp = GPR(y, x, kern)
+
+        n_test = 5
+        x_test = np.random.randn(n_test, dx)
+        f_samples = gp.predict_f_samples(x_test)
+        assert isinstance(f_samples, th.Tensor)
+        assert f_samples.ndimension() == 3  # [sample x n_test x dy]
+        assert f_samples.shape == (1, n_test, dy)
+
+        n_samples = 3
+        f_samples_2 = gp.predict_f_samples(x_test, n_samples=n_samples)
+        assert isinstance(f_samples_2, th.Tensor)
+        assert f_samples_2.ndimension() == 3  # [sample x n_test x dy]
+        assert f_samples_2.shape == (n_samples, n_test, dy)
+
+    def test_predict_y_samples(self):
+        # TODO mock a GPModel?  Using GPR for the moment.
+        n, dx, dy = 5, 3, 2
+        x, y = np.random.randn(n, dx), np.random.randn(n, dy)
+        kern = Rbf(dx, ARD=True)
+        gp = GPR(y, x, kern)
+
+        n_test = 5
+        x_test = np.random.randn(n_test, dx)
+        y_samples = gp.predict_y_samples(x_test)
+        assert isinstance(y_samples, th.Tensor)
+        assert y_samples.ndimension() == 3  # [sample x n_test x dy]
+        assert y_samples.shape == (1, n_test, dy)
+
+        n_samples = 3
+        y_samples_2 = gp.predict_y_samples(x_test, n_samples=n_samples)
+        assert isinstance(y_samples_2, th.Tensor)
+        assert y_samples_2.ndimension() == 3  # [sample x n_test x dy]
+        assert y_samples_2.shape == (n_samples, n_test, dy)

test/test_models/__init__.py

@@ -0,0 +1,3 @@
+# File: __init__.py
+# File Created: Sunday, 21st July 2019 8:06:08 pm
+# Author: Steven Atkinson ([email protected])

test/test_models/test_gpr.py

@@ -0,0 +1,14 @@
+# File: test_gpr.py
+# File Created: Sunday, 21st July 2019 7:53:26 pm
+# Author: Steven Atkinson ([email protected])
+
+"""
+Tests for the GPR class
+"""
+
+import pytest
+
+
+def TestGPR(object):
+    # TODO
+    pass