Coverage for python/lsst/meas/algorithms/reserveSourcesTask.py: 33%

44 statements  

« prev     ^ index     » next       coverage.py v6.4.1, created at 2022-06-23 02:58 -0700

1# 

2# LSST Data Management System 

3# 

4# Copyright 2008-2017 AURA/LSST. 

5# 

6# This product includes software developed by the 

7# LSST Project (http://www.lsst.org/). 

8# 

9# This program is free software: you can redistribute it and/or modify 

10# it under the terms of the GNU General Public License as published by 

11# the Free Software Foundation, either version 3 of the License, or 

12# (at your option) any later version. 

13# 

14# This program is distributed in the hope that it will be useful, 

15# but WITHOUT ANY WARRANTY; without even the implied warranty of 

16# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 

17# GNU General Public License for more details. 

18# 

19# You should have received a copy of the LSST License Statement and 

20# the GNU General Public License along with this program. If not, 

21# see <https://www.lsstcorp.org/LegalNotices/>. 

22# 

23 

24__all__ = ["ReserveSourcesConfig", "ReserveSourcesTask"] 

25 

26import numpy as np 

27 

28from lsst.pex.config import Config, Field 

29from lsst.pipe.base import Task, Struct 

30 

31 

32class ReserveSourcesConfig(Config): 

33 """Configuration for reserving sources""" 

34 fraction = Field(dtype=float, default=0.0, 

35 doc="Fraction of candidates to reserve from fitting; none if <= 0") 

36 seed = Field(dtype=int, default=1, 

37 doc=("This number will be added to the exposure ID to set the random seed for " 

38 "reserving candidates")) 

39 

40 

41class ReserveSourcesTask(Task): 

42 """Reserve sources from analysis 

43 

44 We randomly select a fraction of sources that will be reserved 

45 from analysis. This allows evaluation of the quality of model fits 

46 using sources that were not involved in the fitting process. 

47 

48 Constructor parameters 

49 ---------------------- 

50 columnName : `str`, required 

51 Name of flag column to add; we will suffix this with "_reserved". 

52 schema : `lsst.afw.table.Schema`, required 

53 Catalog schema. 

54 doc : `str` 

55 Documentation for column to add. 

56 config : `ReserveSourcesConfig` 

57 Configuration. 

58 """ 

59 ConfigClass = ReserveSourcesConfig 

60 _DefaultName = "reserveSources" 

61 

62 def __init__(self, columnName=None, schema=None, doc=None, **kwargs): 

63 Task.__init__(self, **kwargs) 

64 assert columnName is not None, "columnName not provided" 

65 assert schema is not None, "schema not provided" 

66 self.columnName = columnName 

67 self.key = schema.addField(self.columnName + "_reserved", type="Flag", doc=doc) 

68 

69 def run(self, sources, prior=None, expId=0): 

70 """Select sources to be reserved 

71 

72 Reserved sources will be flagged in the catalog, and we will return 

73 boolean arrays that identify the sources to be reserved from and 

74 used in the analysis. Typically you'll want to use the sources 

75 from the `use` array in your fitting, and use the sources from the 

76 `reserved` array as an independent test of your fitting. 

77 

78 Parameters 

79 ---------- 

80 sources : `lsst.afw.table.Catalog` or `list` of `lsst.afw.table.Record` 

81 Sources from which to select some to be reserved. 

82 prior : `numpy.ndarray` of type `bool`, optional 

83 Prior selection of sources. Should have the same length as 

84 `sources`. If set, we will only consider for reservation sources 

85 that are flagged `True` in this array. 

86 expId : `int` 

87 Exposure identifier; used for seeding the random number generator. 

88 

89 Return struct contents 

90 ---------------------- 

91 reserved : `numpy.ndarray` of type `bool` 

92 Sources to be reserved are flagged `True` in this array. 

93 use : `numpy.ndarray` of type `bool` 

94 Sources the user should use in analysis are flagged `True`. 

95 """ 

96 if prior is not None: 

97 assert len(prior) == len(sources), "Length mismatch: %s vs %s" % (len(prior), len(sources)) 

98 numSources = prior.sum() 

99 else: 

100 numSources = len(sources) 

101 selection = self.select(numSources, expId) 

102 if prior is not None: 

103 selection = self.applySelectionPrior(prior, selection) 

104 self.markSources(sources, selection) 

105 self.log.info("Reserved %d/%d sources", selection.sum(), len(selection)) 

106 return Struct(reserved=selection, 

107 use=prior & ~selection if prior is not None else np.logical_not(selection)) 

108 

109 def select(self, numSources, expId=0): 

110 """Randomly select some sources 

111 

112 We return a boolean array with a random selection. The fraction 

113 of sources selected is specified by the config parameter `fraction`. 

114 

115 Parameters 

116 ---------- 

117 numSources : `int` 

118 Number of sources in catalog from which to select. 

119 expId : `int` 

120 Exposure identifier; used for seeding the random number generator. 

121 

122 Returns 

123 ------- 

124 selection : `numpy.ndarray` of type `bool` 

125 Selected sources are flagged `True` in this array. 

126 """ 

127 selection = np.zeros(numSources, dtype=bool) 

128 if self.config.fraction <= 0: 

129 return selection 

130 reserve = int(np.round(numSources*self.config.fraction)) 

131 selection[:reserve] = True 

132 rng = np.random.RandomState((self.config.seed + expId) & 0xFFFFFFFF) 

133 rng.shuffle(selection) 

134 return selection 

135 

136 def applySelectionPrior(self, prior, selection): 

137 """Apply selection to full catalog 

138 

139 The `select` method makes a random selection of sources. If those 

140 sources don't represent the full population (because a sub-selection 

141 has already been made), then we need to generate a selection covering 

142 the entire population. 

143 

144 Parameters 

145 ---------- 

146 prior : `numpy.ndarray` of type `bool` 

147 Prior selection of sources, identifying the subset from which the 

148 random selection has been made. 

149 selection : `numpy.ndarray` of type `bool` 

150 Selection of sources in subset identified by `prior`. 

151 

152 Returns 

153 ------- 

154 full : `numpy.ndarray` of type `bool` 

155 Selection applied to full population. 

156 """ 

157 full = np.zeros(len(prior), dtype=bool) 

158 full[prior] = selection 

159 return full 

160 

161 def markSources(self, sources, selection): 

162 """Mark sources in a list or catalog 

163 

164 This requires iterating through the list and setting the flag in 

165 each source individually. Even if the `sources` is a `Catalog` 

166 with contiguous records, it's not currently possible to set a boolean 

167 column (DM-6981) so we need to iterate. 

168 

169 Parameters 

170 ---------- 

171 catalog : `lsst.afw.table.Catalog` or `list` of `lsst.afw.table.Record` 

172 Catalog in which to flag selected sources. 

173 selection : `numpy.ndarray` of type `bool` 

174 Selection of sources to mark. 

175 """ 

176 for src, select in zip(sources, selection): 

177 if select: 

178 src.set(self.key, True)