benchmark
__parse_check_p_n_y(p, n, y)
¶
Parses and checks n
, y
and p
, returns (inferred) n
.
Source code in doubtlab/benchmark.py
5 6 7 8 9 10 11 12 13 14 15 16 17 |
|
calculate_precision_recall_at_k(predicate_df, idx_flip, max_k=100, give_random=False, give_ensemble=True)
¶
Plots precision/recall at k
values for flipped label experiments.
Returns an interactive altair visualisation. Make sure it is installed beforehand.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
predicate_df |
the dataframe with predicates from |
required | |
idx_flip |
array that indicates if labels are wrong |
required | |
max_k |
the maximum value for |
100
|
|
give_random |
plot the "at k" statistics for the randomly selected lower bound |
False
|
|
give_ensemble |
plot the "at k" statistics from the reason ensemble |
True
|
Source code in doubtlab/benchmark.py
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 |
|
flip_labels(y, random_seed=42, n=None, p=None)
¶
Flips subset of labels for benchmarking. Recommended for classification.
Either p
or n
should be given. Returns a tuple (y_out, indicator)
-tuple.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
y |
array of labels |
required | |
random_seed |
random seed |
42
|
|
n |
number of labels to flip |
None
|
|
p |
percentage of labels to flip |
None
|
Usage:
import numpy as np
from doubtlab.benchmark import flip_labels
# Let's pretend these are the actual labels
y = np.random.randint(0, 3, 10000)
# You now have some shuffled labels and an indicator
y_out, indicator = flip_labels(y, n=100)
Source code in doubtlab/benchmark.py
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 |
|
plot_precision_recall_at_k(predicate_df, idx_flip, max_k=100, give_random=True, give_ensemble=True)
¶
Plots precision/recall at k
values for flipped label experiments.
Returns an interactive altair visualisation. Make sure it is installed beforehand.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
predicate_df |
the dataframe with predicates from |
required | |
idx_flip |
array that indicates if labels are wrong |
required | |
max_k |
the maximum value for |
100
|
|
give_random |
plot the "at k" statistics for the randomly selected lower bound |
True
|
|
give_ensemble |
plot the "at k" statistics from the reason ensemble |
True
|
Source code in doubtlab/benchmark.py
154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 |
|
shuffle_labels(y, random_seed=42, n=None, p=None)
¶
Shuffles subset of labels for benchmarking. Recommended for regression.
Either p
or n
should be given. Returns a tuple (y_out, indicator)
-tuple.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
y |
array of labels |
required | |
random_seed |
random seed |
42
|
|
n |
number of labels to flip |
None
|
|
p |
percentage of labels to flip |
None
|
Usage:
import numpy as np
from doubtlab.benchmark import shuffle_labels
# Let's pretend these are the actual labels
y = np.random.normal(0, 1, 10000)
# You now have some shuffled labels and an indicator
y_out, indicator = shuffle_labels(y, n=100)
Source code in doubtlab/benchmark.py
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 |
|