easyvvuq.sampling.grid_sampler

  1"""A grid sampler
  2
  3Useful for e.g. hyperparameter search. The "vary" dict contains the values
  4that must be considered per (hyper)parameter, for instance:
  5
  6    vary = {"x1": [0.0, 0.5, 0.1],
  7            "x2 = [1, 3],
  8            "x3" = [True, False]}
  9
 10The sampler will create a tensor grid using all specified 1D parameter
 11values.
 12"""
 13
 14__author__ = "Wouter Edeling"
 15__copyright__ = """
 16
 17    Copyright 2018 Robin A. Richardson, David W. Wright
 18
 19    This file is part of EasyVVUQ
 20
 21    EasyVVUQ is free software: you can redistribute it and/or modify
 22    it under the terms of the Lesser GNU General Public License as published by
 23    the Free Software Foundation, either version 3 of the License, or
 24    (at your option) any later version.
 25
 26    EasyVVUQ is distributed in the hope that it will be useful,
 27    but WITHOUT ANY WARRANTY; without even the implied warranty of
 28    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 29    Lesser GNU General Public License for more details.
 30
 31    You should have received a copy of the Lesser GNU General Public License
 32    along with this program.  If not, see <https://www.gnu.org/licenses/>.
 33
 34"""
 35__license__ = "LGPL"
 36
 37from itertools import product
 38import numpy as np
 39from .base import BaseSamplingElement  # , Vary
 40
 41
 42class Grid_Sampler(BaseSamplingElement, sampler_name="grid_sampler"):
 43
 44    def __init__(self, vary, count=0):
 45        """
 46        Initialize the grid sampler.
 47
 48        Parameters
 49        ----------
 50        vary : dict, or list of dicts
 51            A dictionary containing all 1D values for each parameter. For instance
 52            vary = {"x1": [0.0, 0.5. 1.0], "x2": [True, False]}. This will
 53            create a 2D tensor product of all (x1, x2) parameter combinations.
 54            If a list of vary dicts is specified, each vary dict will be treated
 55            independently to generate points. These dicts do not have to contain
 56            the same parameters. The tensor product points are stored in the 
 57            'points' list, with one tensor product per vary dict.
 58        count : int, optional
 59            Internal counter used to count the number of samples that have
 60            been executed. The default is 0.
 61
 62        Returns
 63        -------
 64        None.
 65
 66        """
 67        # allways add vary to list, even if only a single dict is specified
 68        if not isinstance(vary, list):
 69            vary = [vary]
 70
 71        self.vary = vary
 72        self.count = count
 73        self.points = []
 74
 75        # make sure all parameters are stored in a list or array, even
 76        # if they have only a single value
 77        for _vary in vary:
 78            for param in _vary.keys():
 79                if not isinstance(_vary[param], list) and not isinstance(_vary[param], np.ndarray):
 80                    vary[param] = [vary[param]]
 81
 82            # use dtype=object to allow for multiple different type (float, boolean etc)
 83            self.points.append(np.array(list(product(*list(_vary.values()))), dtype=object))
 84
 85        # the cumulative sizes of all ensembles generated by the vary dicts
 86        self.cumul_sizes = np.cumsum([points.shape[0] for points in self.points])
 87        # add a zero to the beginning (necessary in __next__ subroutine)
 88        self.cumul_sizes = np.insert(self.cumul_sizes, 0, 0)
 89
 90    def is_finite(self):
 91        return True
 92
 93    def n_samples(self):
 94        """Returns the number of samples in this sampler.
 95        """
 96        # return self.points.shape[0]
 97        return self.cumul_sizes[-1]
 98
 99    def get_param_names(self):
100        """
101        Get the names of all parameters that were varied.
102
103        Returns
104        -------
105        param_names : list
106            List of parameter names.
107
108        """
109        param_names = []
110        for _vary in self.vary:
111            for name in _vary.keys():
112                if not name in param_names:
113                    param_names.append(name)
114        return param_names
115
116    def __next__(self):
117        """
118        Return the next sample from the input distributions.
119
120        Raises
121        ------
122        StopIteration
123            Stop iteration when count >= n_samples.
124
125        Returns
126        -------
127        run_dict : dict
128            A dictionary with the random input samples, e.g.
129            {'x1': 0.5, 'x2': False}.
130
131        """
132        if self.count < self.n_samples():
133            vary_idx = np.where(self.count < self.cumul_sizes[1:])[0][0]
134            run_dict = {}
135            i_par = 0
136            for param_name in self.vary[vary_idx].keys():
137                sample_idx = self.count - self.cumul_sizes[vary_idx]
138                run_dict[param_name] = self.points[vary_idx][sample_idx][i_par]
139                i_par += 1
140            self.count += 1
141            return run_dict
142        else:
143            raise StopIteration