# 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.
# Corresponds to Task T.1 from the paper https://arxiv.org/pdf/2101.04653.pdf
# NOTE: The paper description uses a variance of 0.1 for the prior and likelihood
# but the implementation uses scale = 0.1 Our implmenetation uses a default scale
# of 0.1 for consistency with the implementation.
import numpy as np
bayesflow_benchmark_info = {"simulator_is_batched": True, "parameter_names": None, "configurator_info": "posterior"}
[docs]
def prior(D=10, scale=0.1, rng=None):
"""Generates a random draw from a D-dimensional Gaussian prior distribution with a
spherical scale matrix given by sigma * I_D. Represents the location vector of
a (conjugate) Gaussian likelihood.
Parameters
----------
D : int, optional, default : 10
The dimensionality of the Gaussian prior distribution.
scale : float, optional, default : 0.1
The scale of the Gaussian prior.
rng : np.random.Generator or None, default: None
An optional random number generator to use.
Returns
-------
theta : np.ndarray of shape (D, )
A single draw from the D-dimensional Gaussian prior.
"""
if rng is None:
rng = np.random.default_rng()
return scale * rng.normal(size=D)
[docs]
def simulator(theta, n_obs=None, scale=0.1, rng=None):
"""Generates batched draws from a D-dimenional Gaussian distributions given a batch of
location (mean) parameters of D dimensions. Assumes a spherical convariance matrix given
by scale * I_D.
Parameters
----------
theta : np.ndarray of shape (theta, D)
The location parameters of the Gaussian likelihood.
n_obs : int or None, optional, default: None
The number of observations to draw from the likelihood given the location
parameter `theta`. If `n obs is None`, a single draw is produced.
scale : float, optional, default : 0.1
The scale of the Gaussian likelihood.
rng : np.random.Generator or None, default: None
An optional random number generator to use.
Returns
-------
x : np.ndarray of shape (theta.shape[0], theta.shape[1]) if n_obs is None,
else np.ndarray of shape (theta.shape[0], n_obs, theta.shape[1])
A single draw or a sample from a batch of Gaussians.
"""
# Use default RNG, if None provided
if rng is None:
rng = np.random.default_rng()
# Generate prior predictive samples, possibly a single if n_obs is None
if n_obs is None:
return rng.normal(loc=theta, scale=scale)
x = rng.normal(loc=theta, scale=scale, size=(n_obs, theta.shape[0], theta.shape[1]))
return np.transpose(x, (1, 0, 2))
[docs]
def configurator(forward_dict, mode="posterior"):
"""Configures simulator outputs for use in BayesFlow training."""
# Case only posterior configuration
if mode == "posterior":
input_dict = _config_posterior(forward_dict)
# Case only plikelihood configuration
elif mode == "likelihood":
input_dict = _config_likelihood(forward_dict)
# Case posterior and likelihood configuration (i.e., joint inference)
elif mode == "joint":
input_dict = {}
input_dict["posterior_inputs"] = _config_posterior(forward_dict)
input_dict["likelihood_inputs"] = _config_likelihood(forward_dict)
# Throw otherwise
else:
raise NotImplementedError('For now, only a choice between ["posterior", "likelihood", "joint"] is available!')
return input_dict
def _config_posterior(forward_dict):
"""Helper function for posterior configuration."""
input_dict = {}
input_dict["parameters"] = forward_dict["prior_draws"].astype(np.float32)
input_dict["direct_conditions"] = forward_dict["sim_data"].astype(np.float32)
return input_dict
def _config_likelihood(forward_dict):
"""Helper function for likelihood configuration."""
input_dict = {}
input_dict["conditions"] = forward_dict["prior_draws"].astype(np.float32)
input_dict["observables"] = forward_dict["sim_data"].astype(np.float32)
return input_dict
