# Copyright (c) 2022 The BayesFlow Developers
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
# This module implements all 10 benchmark models (tasks) from the paper:
#
# Lueckmann, J. M., Boelts, J., Greenberg, D., Goncalves, P., & Macke, J. (2021).
# Benchmarking simulation-based inference.
# In International Conference on Artificial Intelligence and Statistics (pp. 343-351). PMLR.
#
# https://arxiv.org/pdf/2101.04653.pdf
#
# However, it lifts the dependency on `PyTorch` and implements the models as ready-made
# tuples of prior and simulator functions capable of interacting with BayesFlow.
# Note: All default hyperparameters are set according to the paper.
import importlib
from functools import partial
import numpy as np
from bayesflow.exceptions import ConfigurationError
from bayesflow.simulation import GenerativeModel, Prior
available_benchmarks = [
"gaussian_linear",
"gaussian_linear_uniform",
"slcp",
"slcp_distractors",
"bernoulli_glm",
"bernoulli_glm_raw",
"gaussian_mixture",
"two_moons",
"sir",
"lotka_volterra",
"inverse_kinematics",
]
[docs]
def get_benchmark_module(benchmark_name):
"""Loads the corresponding benchmark file under bayesflow.benchmarks.<benchmark_name> as a
module and returns it.
"""
try:
benchmark_module = importlib.import_module(f"bayesflow.benchmarks.{benchmark_name}")
return benchmark_module
except ModuleNotFoundError:
raise ConfigurationError(f"You need to provide a valid name from: {available_benchmarks}")
[docs]
class Benchmark:
"""Interface class for a benchmark."""
[docs]
def __init__(self, name, mode="joint", seed=None, **kwargs):
"""Creates a benchmark generative model by using the blueprint contained
in a benchmark file.
Parameters
----------
name : str
The name of the benchmark file (without suffix, i.e., .py) to use as a blueprint.
mode : str, otpional, default: 'joint'
The mode in which to configure the data, should be in ('joint', 'posterior', 'likelihood')
seed : int or None, optional, default: None
The seed to use if reproducibility is required. Will be passed to a numpy RNG.
**kwargs : dict
Optional keyword arguments.
If 'sim_kwargs' is present, key-value pairs will be interpreted as arguments for the simulator
and propagated accordingly.
If 'prior_kwargs' is present, key-value pairs will be interpreted as arguments for the prior
and propagated accordingly.
"""
self.benchmark_name = name
self._rng = np.random.default_rng(seed)
self.benchmark_module = get_benchmark_module(self.benchmark_name)
self.benchmark_info = getattr(self.benchmark_module, "bayesflow_benchmark_info")
# Prepare partial simulator function with optional keyword arguments
if kwargs.get("sim_kwargs") is not None:
_simulator = partial(
getattr(self.benchmark_module, "simulator"), rng=self._rng, **kwargs.pop("sim_kwargs", {})
)
else:
_simulator = partial(getattr(self.benchmark_module, "simulator"), rng=self._rng)
# Prepare partial prior function with optional keyword arguments
if kwargs.get("prior_kwargs") is not None:
_prior = partial(getattr(self.benchmark_module, "prior"), rng=self._rng, **kwargs.pop("prior_kwargs", {}))
else:
_prior = partial(getattr(self.benchmark_module, "prior"), rng=self._rng)
# Prepare generative model
self.generative_model = GenerativeModel(
prior=Prior(
prior_fun=_prior,
param_names=self.benchmark_info["parameter_names"],
),
simulator=_simulator,
simulator_is_batched=self.benchmark_info["simulator_is_batched"],
name=self.benchmark_name,
)
self.configurator = getattr(self.benchmark_module, "configurator")
self.configurator = partial(self.configurator, mode=mode)
