nearly ready to release, lookg at doc

glimix_core/_util/__init__.py

@@ -4,6 +4,7 @@
 from .eigen import economic_qs_zeros
 from .format import format_function
 from .hsolve import hsolve
+from ._assert import assert_interface
 
 log2pi = 1.837877066409345339081937709124758839607238769531250
 
@@ -17,4 +18,5 @@
     "vec",
     "unvec",
     "format_function",
+    "assert_interface",
 ]

glimix_core/_util/_assert.py

@@ -0,0 +1,16 @@
+def assert_interface(cls, req_attrs):
+    attrs = dir(cls)
+    private = set(a for a in attrs if a.startswith("_"))
+    public = set(attrs) - private
+
+    missing = public - set(req_attrs)
+    if missing:
+        msg = "The following public attributes exist but have not been asserted: "
+        msg += "{}".format(", ".join(list(missing)))
+        raise AssertionError(msg)
+
+    missing = set(req_attrs) - public
+    if missing:
+        msg = "The following attributes have not been found: "
+        msg += "{}".format(", ".join(list(missing)))
+        raise AssertionError(msg)

glimix_core/cov/kron2sum.py

@@ -16,9 +16,9 @@
 
 from optimix import Function
 
-from .._util import format_function, unvec
 from .free import FreeFormCov
 from .lrfree import LRFreeFormCov
+from .._util import format_function, unvec
 
 
 class Kron2SumCov(Function):

glimix_core/lmm/_core.py

@@ -216,13 +216,14 @@ def _yTQDiQTm(self):
 
     @property
     def X(self):
-        r"""Covariates set by the user.
+        """
+        Covariates set by the user.
 
         It has to be a matrix of number-of-samples by number-of-covariates.
 
         Returns
         -------
-        :class:`numpy.ndarray`
+        ndarray
             Covariates.
         """
         from numpy_sugar.linalg import ddot

glimix_core/lmm/_rkron2sum.py

@@ -271,7 +271,7 @@ def gradient(self):
         """
         Gradient of the log of the marginal likelihood.
         """
-        return self.lml_gradient()
+        return self._lml_gradient()
 
     @property
     @lru_cache(maxsize=None)
@@ -367,7 +367,7 @@ def lml(self):
 
         return lml / 2
 
-    def lml_gradient(self):
+    def _lml_gradient(self):
         """
         Gradient of the log of the marginal likelihood.
 
@@ -546,7 +546,7 @@ def fit(self, verbose=True):
             ``True`` for progress output; ``False`` otherwise.
             Defaults to ``True``.
         """
-        self._maximize(verbose=verbose)
+        self._maximize(verbose=verbose, pgtol=1e-6)
 
 
 def _dot(a, b):

glimix_core/lmm/test/test_lmm_lmm.py

@@ -2,6 +2,7 @@
 from numpy.random import RandomState
 from numpy.testing import assert_allclose
 
+from glimix_core._util import assert_interface
 from glimix_core.cov import EyeCov, LinearCov, SumCov
 from glimix_core.lik import DeltaProdLik
 from glimix_core.lmm import LMM
@@ -54,3 +55,35 @@ def test_lmm_lmm_prediction():
     K = dot(X, X.T)
     pm = lmm.predictive_mean(ones((n, 1)), K, K.diagonal())
     assert_allclose(corrcoef(y, pm)[0, 1], 0.8358820971891354)
+
+
+def test_lmm_lmm_public_attrs():
+    assert_interface(
+        LMM,
+        [
+            "lml",
+            "X",
+            "beta",
+            "delta",
+            "scale",
+            "mean_star",
+            "variance_star",
+            "covariance_star",
+            "covariance",
+            "predictive_covariance",
+            "mean",
+            "isfixed",
+            "fixed_effects_variance",
+            "gradient",
+            "v0",
+            "v1",
+            "fit",
+            "copy",
+            "value",
+            "get_fast_scanner",
+            "predictive_mean",
+            "name",
+            "unfix",
+            "fix",
+        ],
+    )

glimix_core/lmm/test/test_lmm_rkron2sum.py

@@ -1,10 +1,11 @@
 from numpy.random import RandomState
-from numpy.testing import assert_allclose
+from numpy.testing import assert_allclose, assert_equal
 
+from glimix_core._util import assert_interface
 from glimix_core.lmm import RKron2Sum
 
 
-def test_lmm_reml_rkron2sum():
+def test_lmm_rkron2sum():
     random = RandomState(0)
     n = 5
     Y = random.randn(n, 3)
@@ -15,8 +16,10 @@ def test_lmm_reml_rkron2sum():
     lmm = RKron2Sum(Y, A, F, G)
 
     assert_allclose(lmm.lml(), -16.580821931417656)
-    # assert_allclose(lmm._check_grad(step=1e-7), 0, atol=1e-3)
     assert_allclose(lmm._check_grad(step=1e-7), 0, atol=1e-4)
+    assert_equal(lmm.nsamples, n)
+    assert_equal(lmm.ntraits, 3)
+    assert_equal(lmm.ncovariates, 2)
 
     n = 5
     Y = random.randn(n, 1)
@@ -25,6 +28,53 @@ def test_lmm_reml_rkron2sum():
     F = random.randn(n, 2)
     G = random.randn(n, 4)
     lmm = RKron2Sum(Y, A, F, G)
+    lmm.name = "KronSum"
 
     assert_allclose(lmm.lml(), -4.582089407009583)
     assert_allclose(lmm._check_grad(step=1e-7), 0, atol=1e-4)
+    assert_allclose(
+        [lmm.mean.value()[0], lmm.mean.value()[1]],
+        [0.0497438970225256, 0.5890598193072355],
+    )
+
+    assert_allclose(
+        [
+            lmm.cov.value()[0, 0],
+            lmm.cov.value()[0, 1],
+            lmm.cov.value()[1, 0],
+            lmm.cov.value()[1, 1],
+        ],
+        [
+            4.3712532668348185,
+            -0.07239366121399138,
+            -0.07239366121399138,
+            2.7242131674614862,
+        ],
+    )
+
+    assert_equal(lmm.nsamples, n)
+    assert_equal(lmm.ntraits, 1)
+    assert_equal(lmm.name, "KronSum")
+    lmm.fit(verbose=False)
+    grad = lmm.gradient()
+    assert_allclose(grad["C0.Lu"], [0], atol=1e-5)
+    assert_allclose(grad["C1.Lu"], [0], atol=1e-5)
+    assert_allclose(lmm.lml(), -0.6930197958236421)
+
+
+def test_lmm_rkron2sum_public_attrs():
+    assert_interface(
+        RKron2Sum,
+        [
+            "fit",
+            "ntraits",
+            "lml",
+            "mean",
+            "nsamples",
+            "value",
+            "cov",
+            "ncovariates",
+            "name",
+            "gradient",
+        ],
+    )

glimix_core/lmm/test/test_lmm_scan.py

@@ -3,6 +3,7 @@
 from numpy.random import RandomState
 from numpy.testing import assert_allclose
 
+from glimix_core._util import assert_interface
 from glimix_core.cov import EyeCov, LinearCov, SumCov
 from glimix_core.lik import DeltaProdLik
 from glimix_core.lmm import LMM, FastScanner
@@ -429,3 +430,7 @@ def test_lmm_scan_interface():
 
     with pytest.raises(ValueError):
         FastScanner(y, X, QS, nan)
+
+
+def test_lmm_scan_public_attrs():
+    assert_interface(FastScanner, ["unset_scale", "null_lml", "set_scale", "fast_scan"])