/
ale.py
163 lines (136 loc) · 7 KB
/
ale.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
#!/usr/bin/python
# -*- coding: utf-8 -*-
##
# ale.py: Adaptive likelihood estimation utilities and models.
##
# © 2017, Chris Ferrie (csferrie@gmail.com) and
# Christopher Granade (cgranade@cgranade.com).
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
#
# 1. Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
#
# 2. Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
#
# 3. Neither the name of the copyright holder nor the names of its
# contributors may be used to endorse or promote products derived from
# this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
# ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
# POSSIBILITY OF SUCH DAMAGE.
##
## FEATURES ###################################################################
from __future__ import absolute_import
from __future__ import division
## ALL ########################################################################
# We use __all__ to restrict what globals are visible to external modules.
__all__ = [
'binom_est_p',
'ALEApproximateModel'
]
## IMPORTS ####################################################################
from itertools import count
import warnings
import numpy as np
from scipy.stats.distributions import binom
from qinfer.utils import binom_est_error, binom_est_p
from qinfer.derived_models import DerivedModel
from qinfer.abstract_model import Model, Simulatable, FiniteOutcomeModel
from qinfer._exceptions import ApproximationWarning
from qinfer._due import due, Doi
## CLASSES ####################################################################
class ALEApproximateModel(DerivedModel):
r"""
Given a :class:`~qinfer.abstract_model.Simulatable`, estimates the
likelihood of that simulator by using adaptive likelihood estimation (ALE).
:param qinfer.abstract_model.Simulatable simulator: Simulator to estimate
the likelihood function of.
:param float error_tol: Allowed error in the estimated likelihood. Note that
the simulation cost scales as :math:`O(\epsilon^{-2})`, where
:math:`\epsilon` is the error tolerance.
:param int min_samp: Minimum number of samples to use in estimating the
likelihood.
:param int samp_step: Number of samples by which to increment if the error
tolerance is not met.
:param float est_hedge: Amount of hedging to use in reporting the final
estimate.
:param float adapt_hedge: Amount of hedging to use in deciding if the error
tolerance has been met. Increasing this parameter will in general
cause the algorithm to require more samples.
"""
@due.dcite(
Doi("10.1103/PhysRevLett.112.130402"),
description="Adaptive likelihood estimation",
tags=["implementation"]
)
def __init__(self, simulator,
error_tol=1e-2, min_samp=10, samp_step=10,
est_hedge=0.509, adapt_hedge=0.509
):
## INPUT VALIDATION ##
if not isinstance(simulator, Simulatable):
raise TypeError("Simulator must be an instance of Simulatable.")
if error_tol <= 0:
raise ValueError("Error tolerance must be strictly positive.")
if error_tol > 1:
raise ValueError("Error tolerance must be less than 1.")
if min_samp <= 0:
raise ValueError("Minimum number of samples (min_samp) must be positive.")
if samp_step <= 0:
raise ValueError("Sample step (samp_step) must be positive.")
if est_hedge < 0:
raise ValueError("Estimator hedging (est_hedge) must be non-negative.")
if adapt_hedge < 0:
raise ValueError("Adaptive hedging (adapt_hedge) must be non-negative.")
# this simulator constraint makes implementation easier
if not (simulator.is_n_outcomes_constant and simulator.n_outcomes(None) == 2):
raise ValueError("Decorated model must be a two-outcome model.")
# We had to have the simulator in place before we could call
# the superclass.
super(ALEApproximateModel, self).__init__(simulator)
self._error_tol = float(error_tol)
self._min_samp = int(min_samp)
self._samp_step = int(samp_step)
self._est_hedge = float(est_hedge)
self._adapt_hedge = float(adapt_hedge)
## WRAPPED METHODS AND PROPERTIES ##
# We only need to wrap sim_count and simulate_experiment,
# since the rest are handled by DerivedModel.
@property
def sim_count(self): return self.underlying_model.sim_count
def simulate_experiment(self, modelparams, expparams, repeat=1):
return self.underlying_model.simulate_experiment(modelparams, expparams, repeat=repeat)
## IMPLEMENTATIONS OF MODEL METHODS ##
def likelihood(self, outcomes, modelparams, expparams):
# FIXME: at present, will proceed until ALL model experiment pairs
# are below error tol.
# Should disable one-by-one, but that's tricky.
super(ALEApproximateModel, self).likelihood(outcomes, modelparams, expparams)
simulator = self.underlying_model
# We will use the fact we have assumed a two-outcome model to make the
# problem easier. As such, we will rely on the static method
# FiniteOutcomeModel.pr0_to_likelihood_array.
# Start off with min_samp samples.
n = np.zeros((modelparams.shape[0], expparams.shape[0]))
for N in count(start=self._min_samp, step=self._samp_step):
sim_data = simulator.simulate_experiment(
modelparams, expparams, repeat=self._samp_step
)
n += np.sum(sim_data, axis=0) # Sum over the outcomes axis to find the
# number of 1s.
error_est_p1 = binom_est_error(binom_est_p(n, N, self._adapt_hedge), N, self._adapt_hedge)
if np.all(error_est_p1 < self._error_tol): break
return FiniteOutcomeModel.pr0_to_likelihood_array(outcomes, 1 - binom_est_p(n, N, self._est_hedge))