Add an interface and basic implementations for truncated variables

Author: Ricardo

aeppl/truncation.py

@@ -0,0 +1,265 @@
+from functools import singledispatch
+from typing import Optional, Tuple
+
+import aesara.tensor as at
+import aesara.tensor.random.basic as arb
+import numpy as np
+from aesara import scan
+from aesara.compile.builders import OpFromGraph
+from aesara.graph.op import Op
+from aesara.raise_op import CheckAndRaise
+from aesara.scan import until
+from aesara.tensor.random import RandomStream
+from aesara.tensor.random.op import RandomVariable
+from aesara.tensor.var import TensorConstant, TensorVariable
+
+from aeppl.abstract import MeasurableVariable, _get_measurable_outputs
+from aeppl.logprob import (
+    CheckParameterValue,
+    _logcdf,
+    _logprob,
+    icdf,
+    logcdf,
+    logdiffexp,
+)
+
+
+class TruncatedRV(OpFromGraph):
+    """An `Op` constructed from an Aesara graph that represents a truncated univariate random variable."""
+
+    default_output = 0
+    base_rv_op = None
+
+    def __init__(self, base_rv_op: Op, *args, **kwargs):
+        self.base_rv_op = base_rv_op
+        super().__init__(*args, **kwargs)
+
+
+MeasurableVariable.register(TruncatedRV)
+
+
+@_get_measurable_outputs.register(TruncatedRV)
+def _get_measurable_outputs_TruncatedRV(op, node):
+    return [node.outputs[0]]
+
+
+@singledispatch
+def _truncated(op: Op, lower, upper, *params):
+    """Return the truncated equivalent of another `RandomVariable`."""
+    raise NotImplementedError(
+        f"{op} does not have an equivalent truncated version implemented"
+    )
+
+
+class TruncationError(Exception):
+    """Exception for errors generated from truncated graphs"""
+
+
+class TruncationCheck(CheckAndRaise):
+    """Implements a check in truncated graphs.
+
+    Raises `TruncationError` if the check is not True.
+    """
+
+    def __init__(self, msg=""):
+        super().__init__(TruncationError, msg)
+
+    def __str__(self):
+        return f"TruncationCheck{{{self.msg}}}"
+
+
+def truncate(
+    rv: TensorVariable,
+    lower=None,
+    upper=None,
+    max_n_steps: int = 10_000,
+    srng: Optional[RandomStream] = None,
+) -> Tuple[TensorVariable, Tuple[TensorVariable, TensorVariable]]:
+    """Truncate a univariate `RandomVariable` between `lower` and `upper`.
+
+    If `lower` or `upper` is ``None``, the variable is not truncated on that side.
+
+    Depending on whether or not a dispatch implementation is available, this
+    function returns either a specialized `Op`, or an equivalent graph
+    representing the truncation process via inverse CDF or rejection
+    sampling.
+
+    The argument `max_n_steps` controls the maximum number of resamples that are
+    attempted when performing rejection sampling. A `TruncationError` is raised if
+    convergence is not reached after that many steps.
+
+    Returns
+    =======
+    `TensorVariable` graph representing the truncated `RandomVariable` and respective updates
+    """
+
+    if lower is None and upper is None:
+        raise ValueError("lower and upper cannot both be None")
+
+    if not (isinstance(rv.owner.op, RandomVariable) and rv.owner.op.ndim_supp == 0):
+        raise NotImplementedError(
+            f"Truncation is only implemented for univariate random variables, got {rv.owner.op}"
+        )
+
+    lower = at.as_tensor_variable(lower) if lower is not None else at.constant(-np.inf)
+    upper = at.as_tensor_variable(upper) if upper is not None else at.constant(np.inf)
+
+    if srng is None:
+        srng = RandomStream()
+
+    # Try to use specialized Op
+    try:
+        truncated_rv, updates = _truncated(
+            rv.owner.op, lower, upper, srng, *rv.owner.inputs[1:]
+        )
+        return truncated_rv, updates
+    except NotImplementedError:
+        pass
+
+    # Variables with `_` suffix identify dummy inputs for the OpFromGraph
+    # We will use the Shared RNG variable directly because Scan demands it, even
+    # though it would not be necessary for the icdf OpFromGraph
+    graph_inputs = [*rv.owner.inputs[1:], lower, upper]
+    graph_inputs_ = [inp.type() for inp in graph_inputs]
+    size_, dtype_, *rv_inputs_, lower_, upper_ = graph_inputs_
+    rv_ = srng.gen(rv.owner.op, *rv_inputs_, size=size_, dtype=dtype_)
+
+    # Try to use inverted cdf sampling
+    try:
+        # For left truncated discrete RVs, we need to include the whole lower bound.
+        # This may result in draws below the truncation range, if any uniform == 0
+        lower_value = lower_ - 1 if rv.owner.op.dtype.startswith("int") else lower_
+        cdf_lower_ = at.exp(logcdf(rv_, lower_value))
+        cdf_upper_ = at.exp(logcdf(rv_, upper_))
+        uniform_ = srng.uniform(
+            cdf_lower_,
+            cdf_upper_,
+            size=size_,
+        )
+        truncated_rv_ = icdf(rv_, uniform_)
+        truncated_rv = TruncatedRV(
+            base_rv_op=rv.owner.op,
+            inputs=graph_inputs_,
+            outputs=[truncated_rv_, uniform_.owner.outputs[0]],
+            inline=True,
+        )(*graph_inputs)
+        updates = {truncated_rv.owner.inputs[-1]: truncated_rv.owner.outputs[-1]}
+        return truncated_rv, updates
+    except NotImplementedError:
+        pass
+
+    # Fallback to rejection sampling
+    # TODO: Handle potential broadcast by lower / upper
+    def loop_fn(truncated_rv, reject_draws, lower, upper, size, dtype, *rv_inputs):
+        new_truncated_rv = srng.gen(rv.owner.op, *rv_inputs, size=size, dtype=dtype)  # type: ignore
+        truncated_rv = at.set_subtensor(
+            truncated_rv[reject_draws],
+            new_truncated_rv[reject_draws],
+        )
+        reject_draws = at.or_((truncated_rv < lower), (truncated_rv > upper))
+
+        return (truncated_rv, reject_draws), until(~at.any(reject_draws))
+
+    (truncated_rv_, reject_draws_), updates = scan(
+        loop_fn,
+        outputs_info=[
+            at.zeros_like(rv_),
+            at.ones_like(rv_, dtype=bool),
+        ],
+        non_sequences=[lower_, upper_, size_, dtype_, *rv_inputs_],
+        n_steps=max_n_steps,
+        strict=True,
+    )
+
+    truncated_rv_ = truncated_rv_[-1]
+    convergence_ = ~at.any(reject_draws_[-1])
+    truncated_rv_ = TruncationCheck(
+        f"Truncation did not converge in {max_n_steps} steps"
+    )(truncated_rv_, convergence_)
+
+    truncated_rv = TruncatedRV(
+        base_rv_op=rv.owner.op,
+        inputs=graph_inputs_,
+        # This will fail with `n_steps==1`, because in that case `Scan` won't return any updates
+        outputs=[truncated_rv_, rv_.owner.outputs[0], tuple(updates.values())[0]],
+        inline=True,
+    )(*graph_inputs)
+    # TODO: Is the order of multiple shared variables determnistic?
+    assert truncated_rv.owner.inputs[-2] is rv_.owner.inputs[0]
+    updates = {
+        truncated_rv.owner.inputs[-2]: truncated_rv.owner.outputs[-2],
+        truncated_rv.owner.inputs[-1]: truncated_rv.owner.outputs[-1],
+    }
+    return truncated_rv, updates
+
+
+@_logprob.register(TruncatedRV)
+def truncated_logprob(op, values, *inputs, **kwargs):
+    (value,) = values
+
+    # Rejection sample graph has two rngs
+    if len(op.shared_inputs) == 2:
+        *rv_inputs, lower_bound, upper_bound, _, rng = inputs
+    else:
+        *rv_inputs, lower_bound, upper_bound, rng = inputs
+    rv_inputs = [rng, *rv_inputs]
+
+    base_rv_op = op.base_rv_op
+    logp = _logprob(base_rv_op, (value,), *rv_inputs, **kwargs)
+    # For left truncated RVs, we don't want to include the lower bound in the
+    # normalization term
+    lower_bound_value = (
+        lower_bound - 1 if base_rv_op.dtype.startswith("int") else lower_bound
+    )
+    lower_logcdf = _logcdf(base_rv_op, lower_bound_value, *rv_inputs, **kwargs)
+    upper_logcdf = _logcdf(base_rv_op, upper_bound, *rv_inputs, **kwargs)
+
+    if base_rv_op.name:
+        logp.name = f"{base_rv_op}_logprob"
+        lower_logcdf.name = f"{base_rv_op}_lower_logcdf"
+        upper_logcdf.name = f"{base_rv_op}_upper_logcdf"
+
+    is_lower_bounded = not (
+        isinstance(lower_bound, TensorConstant)
+        and np.all(np.isneginf(lower_bound.value))
+    )
+    is_upper_bounded = not (
+        isinstance(upper_bound, TensorConstant) and np.all(np.isinf(upper_bound.value))
+    )
+
+    lognorm = 0
+    if is_lower_bounded and is_upper_bounded:
+        lognorm = logdiffexp(upper_logcdf, lower_logcdf)
+    elif is_lower_bounded:
+        lognorm = at.log1mexp(lower_logcdf)
+    elif is_upper_bounded:
+        lognorm = upper_logcdf
+
+    logp = logp - lognorm
+
+    if is_lower_bounded:
+        logp = at.switch(value < lower_bound, -np.inf, logp)
+
+    if is_upper_bounded:
+        logp = at.switch(value <= upper_bound, logp, -np.inf)
+
+    if is_lower_bounded and is_upper_bounded:
+        logp = CheckParameterValue("lower_bound <= upper_bound")(
+            logp, at.all(at.le(lower_bound, upper_bound))
+        )
+
+    return logp
+
+
+@_truncated.register(arb.UniformRV)
+def uniform_truncated(op, lower, upper, srng, size, dtype, lower_orig, upper_orig):
+    truncated_uniform = srng.gen(
+        op,
+        at.max((lower_orig, lower)),
+        at.min((upper_orig, upper)),
+        size=size,
+        dtype=dtype,
+    )
+    return truncated_uniform, {
+        truncated_uniform.owner.inputs[0]: truncated_uniform.owner.outputs[0]
+    }