Fixes in logprob derivation of min

Author: Dhruvanshu-Joshi

pymc/logprob/order.py

@@ -41,6 +41,8 @@
 from pytensor.graph.basic import Node
 from pytensor.graph.fg import FunctionGraph
 from pytensor.graph.rewriting.basic import node_rewriter
+from pytensor.scalar.basic import Mul
+from pytensor.tensor.elemwise import Elemwise
 from pytensor.tensor.math import Max
 from pytensor.tensor.random.op import RandomVariable
 from pytensor.tensor.var import TensorVariable
@@ -122,7 +124,94 @@ def max_logprob(op, values, base_rv, **kwargs):
     logcdf = _logcdf_helper(base_rv, value)
 
     [n] = constant_fold([base_rv.size])
-
     logprob = (n - 1) * logcdf + logprob + pt.math.log(n)
 
     return logprob
+
+
+class MeasurableMin(Max):
+    """A placeholder used to specify a log-likelihood for a min sub-graph."""
+
+
+MeasurableVariable.register(MeasurableMin)
+
+
+@node_rewriter(tracks=[Max])
+def find_measurable_min(fgraph: FunctionGraph, node: Node) -> Optional[List[TensorVariable]]:
+    rv_map_feature = getattr(fgraph, "preserve_rv_mappings", None)
+
+    if rv_map_feature is None:
+        return None  # pragma: no cover
+
+    if isinstance(node.op, MeasurableMin):
+        return None  # pragma: no cover
+
+    base_var = node.inputs[0]
+
+    if base_var.owner is None:
+        return None
+
+    if not rv_map_feature.request_measurable(node.inputs):
+        return None
+
+    # Min is the Max of the negation of the same distribution. Hence, op must be Elemiwise
+    if not isinstance(base_var.owner.op, Elemwise):
+        return None
+
+    # negation is -1*(rv). Hence the scalar_op must be Mul
+    if not isinstance(base_var.owner.op.scalar_op, Mul):
+        return None
+
+    base_rv = base_var.owner.inputs[0]
+
+    # Non-univariate distributions and non-RVs must be rejected
+    if not (isinstance(base_rv.owner.op, RandomVariable) and base_rv.owner.op.ndim_supp == 0):
+        return None
+
+    # TODO: We are currently only supporting continuous rvs
+    if isinstance(base_rv.owner.op, RandomVariable) and base_rv.owner.op.dtype.startswith("int"):
+        return None
+
+    # univariate i.i.d. test which also rules out other distributions
+    for params in base_rv.owner.inputs[3:]:
+        if params.type.ndim != 0:
+            return None
+
+    # Check whether axis is supported or not
+    axis = set(node.op.axis)
+    base_var_dims = set(range(base_var.ndim))
+    if axis != base_var_dims:
+        return None
+
+    measurable_min = MeasurableMin(list(axis))
+    min_rv_node = measurable_min.make_node(base_var)
+    min_rv = min_rv_node.outputs
+
+    return min_rv
+
+
+measurable_ir_rewrites_db.register(
+    "find_measurable_min",
+    find_measurable_min,
+    "basic",
+    "min",
+)
+
+
+@_logprob.register(MeasurableMin)
+def min_logprob(op, values, base_var, **kwargs):
+    r"""Compute the log-likelihood graph for the `Max` operation.
+    The formula that we use here is :
+        \ln(f_{(n)}(x)) = \ln(n) + (n-1) \ln(1 - F(x)) + \ln(f(x))
+    where f(x) represents the p.d.f and F(x) represents the c.d.f of the distribution respectively.
+    """
+    (value,) = values
+    base_rv = base_var.owner.inputs[0]
+
+    logprob = _logprob_helper(base_rv, value)
+    logcdf = _logcdf_helper(base_rv, value)
+
+    [n] = constant_fold([base_rv.size])
+    logprob = (n - 1) * pt.math.log(1 - pt.math.exp(logcdf)) + logprob + pt.math.log(n)
+
+    return logprob

tests/logprob/test_order.py

@@ -43,7 +43,6 @@
 import pymc as pm
 
 from pymc import logp
-from pymc.logprob import conditional_logp
 from pymc.testing import assert_no_rvs
 
 
@@ -58,55 +57,112 @@ def test_argmax():
         x_max_logprob = logp(x_max, x_max_value)
 
 
-def test_max_non_iid_fails():
-    """Test whether the logprob for ```pt.max``` for non i.i.d is correctly rejected"""
+@pytest.mark.parametrize(
+    "if_max",
+    [
+        True,
+        False,
+    ],
+)
+def test_non_iid_fails(if_max):
+    """Test whether the logprob for ```pt.max``` or ```pt.min``` for non i.i.d is correctly rejected"""
     x = pm.Normal.dist([0, 1, 2, 3, 4], 1, shape=(5,))
     x.name = "x"
-    x_max = pt.max(x, axis=-1)
-    x_max_value = pt.vector("x_max_value")
+    if if_max == True:
+        x_m = pt.max(x, axis=-1)
+        x_m_value = pt.vector("x_max_value")
+    else:
+        x_min = pt.min(x, axis=-1)
+        x_m = x_min.owner.inputs[0]
+        x_m_value = pt.vector("x_min_value")
     with pytest.raises(RuntimeError, match=re.escape("Logprob method not implemented")):
-        x_max_logprob = logp(x_max, x_max_value)
+        x_max_logprob = logp(x_m, x_m_value)
 
 
-def test_max_non_rv_fails():
+@pytest.mark.parametrize(
+    "if_max",
+    [True, False],
+)
+def test_non_rv_fails(if_max):
     """Test whether the logprob for ```pt.max``` for non-RVs is correctly rejected"""
     x = pt.exp(pt.random.beta(0, 1, size=(3,)))
     x.name = "x"
-    x_max = pt.max(x, axis=-1)
-    x_max_value = pt.vector("x_max_value")
+    if if_max == True:
+        x_m = pt.max(x, axis=-1)
+        x_m_value = pt.vector("x_max_value")
+    else:
+        x_min = pt.min(x, axis=-1)
+        x_m = x_min.owner.inputs[0]
+        x_m_value = pt.vector("x_min_value")
     with pytest.raises(RuntimeError, match=re.escape("Logprob method not implemented")):
-        x_max_logprob = logp(x_max, x_max_value)
+        x_max_logprob = logp(x_m, x_m_value)
 
 
-def test_max_multivariate_rv_fails():
+@pytest.mark.parametrize(
+    "if_max",
+    [
+        True,
+        False,
+    ],
+)
+def test_multivariate_rv_fails(if_max):
     _alpha = pt.scalar()
     _k = pt.iscalar()
     x = pm.StickBreakingWeights.dist(_alpha, _k)
     x.name = "x"
-    x_max = pt.max(x, axis=-1)
-    x_max_value = pt.vector("x_max_value")
+    if if_max == True:
+        x_m = pt.max(x, axis=-1)
+        x_m_value = pt.vector("x_max_value")
+    else:
+        x_min = pt.min(x, axis=-1)
+        x_m = x_min.owner.inputs[0]
+        x_m_value = pt.vector("x_min_value")
     with pytest.raises(RuntimeError, match=re.escape("Logprob method not implemented")):
-        x_max_logprob = logp(x_max, x_max_value)
+        x_max_logprob = logp(x_m, x_m_value)
 
 
-def test_max_categorical():
+@pytest.mark.parametrize(
+    "if_max",
+    [
+        True,
+        False,
+    ],
+)
+def test_categorical(if_max):
     """Test whether the logprob for ```pt.max``` for unsupported distributions is correctly rejected"""
     x = pm.Categorical.dist([1, 1, 1, 1], shape=(5,))
     x.name = "x"
-    x_max = pt.max(x, axis=-1)
-    x_max_value = pt.vector("x_max_value")
+    if if_max == True:
+        x_m = pt.max(x, axis=-1)
+        x_m_value = pt.vector("x_max_value")
+    else:
+        x_min = pt.min(x, axis=-1)
+        x_m = x_min.owner.inputs[0]
+        x_m_value = pt.vector("x_min_value")
     with pytest.raises(RuntimeError, match=re.escape("Logprob method not implemented")):
-        x_max_logprob = logp(x_max, x_max_value)
+        x_max_logprob = logp(x_m, x_m_value)
 
 
-def test_non_supp_axis_max():
+@pytest.mark.parametrize(
+    "if_max",
+    [
+        True,
+        False,
+    ],
+)
+def test_non_supp_axis(if_max):
     """Test whether the logprob for ```pt.max``` for unsupported axis is correctly rejected"""
     x = pt.random.normal(0, 1, size=(3, 3))
     x.name = "x"
-    x_max = pt.max(x, axis=-1)
-    x_max_value = pt.vector("x_max_value")
+    if if_max == True:
+        x_m = pt.max(x, axis=-1)
+        x_m_value = pt.vector("x_max_value")
+    else:
+        x_min = pt.min(x, axis=-1)
+        x_m = x_min.owner.inputs[0]
+        x_m_value = pt.vector("x_min_value")
     with pytest.raises(RuntimeError, match=re.escape("Logprob method not implemented")):
-        x_max_logprob = logp(x_max, x_max_value)
+        x_max_logprob = logp(x_m, x_m_value)
 
 
 @pytest.mark.parametrize(
@@ -147,3 +203,54 @@ def test_max_logprob(shape, value, axis):
         (x_max_logprob.eval({x_max_value: test_value})),
         rtol=1e-06,
     )
+
+
+@pytest.mark.parametrize(
+    "shape, value, axis",
+    [
+        (3, 0.85, -1),
+        (3, 0.01, 0),
+        (2, 0.2, None),
+        (4, 0.5, 0),
+        ((3, 4), 0.9, None),
+        ((3, 4), 0.75, (1, 0)),
+    ],
+)
+def test_min_logprob(shape, value, axis):
+    """Test whether the logprob for ```pt.mix``` produces the corrected
+    The fact that order statistics of i.i.d. uniform RVs ~ Beta is used here:
+        U_1, \\dots, U_n \\stackrel{\text{i.i.d.}}{\\sim} \text{Uniform}(0, 1) \\Rightarrow U_{(k)} \\sim \text{Beta}(k, n + 1- k)
+    for all 1<=k<=n
+    """
+    x = pt.random.uniform(0, 1, size=shape)
+    x.name = "x"
+    x_min = pt.min(x, axis=axis)
+    x_min_rv = x_min.owner.inputs[0]
+    x_min_value = pt.scalar("x_min_value")
+    x_min_logprob = logp(x_min_rv, x_min_value)
+
+    assert_no_rvs(x_min_logprob)
+
+    test_value = value
+
+    n = np.prod(shape)
+    beta_rv = pt.random.beta(1, n, name="beta")
+    beta_vv = beta_rv.clone()
+    beta_rv_logprob = logp(beta_rv, beta_vv)
+
+    np.testing.assert_allclose(
+        beta_rv_logprob.eval({beta_vv: test_value}),
+        (x_min_logprob.eval({x_min_value: test_value})),
+        rtol=1e-06,
+    )
+
+
+def test_min_non_mul_elemwise_fails():
+    """Test whether the logprob for ```pt.min``` for non-mul elemwise RVs is rejected correctly"""
+    x = pt.log(pt.random.beta(0, 1, size=(3,)))
+    x.name = "x"
+    x_min = pt.min(x, axis=-1)
+    x_min_rv = x_min.owner.inputs[0]
+    x_min_value = pt.vector("x_min_value")
+    with pytest.raises(RuntimeError, match=re.escape("Logprob method not implemented")):
+        x_min_logprob = logp(x_min_rv, x_min_value)