Coverage for tadkit/catalog/learners/_confiance_components/_sbad_wrapper.py: 18%

1from typing import List 

2 

3 

4def get_wrapped_dilanodetectm(): 

5 """Returns the TADlearner wrapped from the DiLAnoDetectm method 

6 of the Sparsity-Based Anomaly Detection framework. 

7 

8 The function is intended for use if the dependency is available. 

9 """ 

10 

11 from sbad_fnn.models import DiLAnoDetectm 

12 

13 class DiLAnoDetectmWrapper(DiLAnoDetectm): 

14 

15 def __init__( 

16 self, 

17 nb_atoms: int, 

18 nb_blocks: int, 

19 kernel_size: int, 

20 nb_scales: int, 

21 overlap_perc: float, 

22 layers_sizes: List[int] = None, 

23 activation: str = "id", 

24 wvlt_name: str = "db2", 

25 same: bool = False, 

26 share_weights: bool = False, 

27 soft: bool = False, 

28 shrink: bool = True, 

29 b: float = 1.5, 

30 min_nbsegments: int = 50, 

31 min_lag: int = 64, 

32 ) -> None: 

33 if layers_sizes is None: 

34 layers_sizes = [39, 3] 

35 super().__init__( 

36 nb_atoms=nb_atoms, 

37 nb_blocks=nb_blocks, 

38 kernel_size=kernel_size, 

39 layers_sizes=layers_sizes, 

40 nb_scales=nb_scales, 

41 overlap_perc=overlap_perc, 

42 activation=activation, 

43 wvlt_name=wvlt_name, 

44 same=same, 

45 share_weights=share_weights, 

46 soft=soft, 

47 shrink=shrink, 

48 b=b, 

49 min_nbsegments=min_nbsegments, 

50 min_lag=min_lag, 

51 ) # to be removed when lists are supported by widgets 

52 

53 def fit(self, x): 

54 x = [[x.swapaxes(0, 1)]] 

55 return super().fit(x) 

56 

57 def score_samples(self, x): 

58 x = [[x.swapaxes(0, 1)]] 

59 return -super().score_samples(x)[0][0][0][: len(x)] 

60 

61 DiLAnoDetectmWrapper.required_properties = ["multiple_time_series"] 

62 DiLAnoDetectmWrapper.params_description = { 

63 "nb_atoms": { 

64 "description": "Number of patterns to learn.", 

65 "value_type": "range", 

66 "start": 1, 

67 "step": 1, 

68 "stop": 100, # @martin: stop put to 100 without any info 

69 "default": 2, 

70 }, 

71 "nb_blocks": { 

72 "description": "Number of unrolled gradient steps blocks.", 

73 "value_type": "range", 

74 "start": 1, 

75 "step": 1, 

76 "stop": 100, # @martin: stop put to 100 without any info 

77 "default": 1, 

78 }, 

79 "kernel_size": { 

80 "description": "Convolutional patterns size, must be odd.", 

81 "value_type": "range", 

82 "start": 1, 

83 "step": 2, 

84 "stop": 100, # @martin: stop put to 100 without any info 

85 "default": 5, 

86 }, 

87 # "layers_sizes": { 

88 # "description": "Spectral autoencoder layers sizes.", 

89 # "value_type": "array", 

90 # "length": 2, 

91 # "elements": { 

92 # "value_type": "range", "start": 1, "step": 1, 

93 # }, 

94 # "default": (39, 3), 

95 # }, 

96 "nb_scales": { 

97 "description": "Number of wavelet scales for decomposing time series.", 

98 "value_type": "range", 

99 "start": 1, 

100 "step": 1, 

101 "stop": 100, # @martin: stop put to 100 without any info 

102 "default": 1, 

103 }, 

104 "overlap_perc": { 

105 "description": "Overlapping ratio between sliding windows.", 

106 "value_type": "range", 

107 "start": 0, 

108 "step": 0.01, 

109 "stop": 1, 

110 "default": 0.98, 

111 }, 

112 "same": { 

113 "description": "Whether the blocks are identical or have different weights.", 

114 "value_type": "bool_choice", 

115 "default": False, 

116 }, 

117 "share_weights": { 

118 "description": "Whether within a given block, the conv and conv_transpose layers share weights or not.", 

119 "value_type": "bool_choice", 

120 "default": False, 

121 }, 

122 "soft": { 

123 "description": "Whether the shrinking is soft.", 

124 "value_type": "bool_choice", 

125 "default": False, 

126 }, 

127 "shrink": { 

128 "description": "Whether a shrinking activation function is applied to each block's output.", 

129 "value_type": "bool_choice", 

130 "default": True, 

131 }, 

132 "min_nbsegments": { 

133 "description": "Minimum number of segments.", 

134 "value_type": "range", 

135 "start": 1, 

136 "step": 1, 

137 "stop": 100, # @martin: stop put to 100 without any info 

138 "default": 1, 

139 }, 

140 "min_lag": { 

141 "description": "Minimum lag.", 

142 "value_type": "range", 

143 "start": 1, 

144 "step": 1, 

145 "stop": 100, # @martin: stop put to 100 without any info 

146 "default": 2, 

147 }, 

148 } 

149 

150 return DiLAnoDetectmWrapper