lapros 0.0.1

lapros data for better AI

LaPros

LaPros stands for label probability and label problems

Install

pip install lapros

How to use

The first version of LaPros will rank the suspicious labels given some trustworthy probabilies. The required inputs are a list of observed labels and a 2-dimension list of probabilities. You can use normal Python lists, Numpy arrays or Pandas Series and DataFrames. Return values are in a Numpy array or a Pandas series, the larger the value, the more suspicious are the coresponding labels.

from lapros.noise import suspect

suspect(
    [1, 0, 0, 1, 1],
    probas=[0.5, 0.6, 0.7, 0.8, 0.9],
)

2022-07-11 12:33:41.868 | DEBUG    | lapros.noise:suspect:77 - Normal Python lists
2022-07-11 12:33:41.869 | DEBUG    | lapros.noise:__rank_suspicious:22 - Numpy arrays labels and probas must have same length 5 vs 5
2022-07-11 12:33:41.869 | DEBUG    | lapros.noise:__rank_suspicious:25 - Unique labels [0 1]
2022-07-11 12:33:41.870 | DEBUG    | lapros.noise:__rank_suspicious:27 - Trying to reshape probas
2022-07-11 12:33:41.870 | DEBUG    | lapros.noise:__rank_suspicious:29 - Shape of labels and probas: (5,) vs (5, 1)
2022-07-11 12:33:41.871 | DEBUG    | lapros.noise:__rank_suspicious:37 - ranks [0.5 0.6 0.7 0.2 0.1]





[array([0.5, 0.6, 0.7, 0.2, 0.1])]

suspect(
    [1, 0, 0, 1, 1],
    probas=[[0.5, 0.6, 0.7, 0.8, 0.9]],
)

2022-07-11 12:33:41.937 | DEBUG    | lapros.noise:suspect:77 - Normal Python lists
2022-07-11 12:33:41.938 | DEBUG    | lapros.noise:__rank_suspicious:22 - Numpy arrays labels and probas must have same length 5 vs 1
2022-07-11 12:33:41.939 | DEBUG    | lapros.noise:__rank_suspicious:25 - Unique labels [0 1]
2022-07-11 12:33:41.940 | DEBUG    | lapros.noise:__rank_suspicious:27 - Trying to reshape probas
2022-07-11 12:33:41.940 | DEBUG    | lapros.noise:__rank_suspicious:29 - Shape of labels and probas: (5,) vs (5, 1)
2022-07-11 12:33:41.940 | DEBUG    | lapros.noise:__rank_suspicious:37 - ranks [0.5 0.6 0.7 0.2 0.1]





[array([0.5, 0.6, 0.7, 0.2, 0.1])]

suspect(
    np.array([1, 0, 0, 1, 1]),
    probas=np.array([0.5, 0.6, 0.7, 0.8, 0.9]),
)

2022-07-11 12:33:42.004 | DEBUG    | lapros.noise:__rank_suspicious:22 - Numpy arrays labels and probas must have same length 5 vs 5
2022-07-11 12:33:42.006 | DEBUG    | lapros.noise:__rank_suspicious:25 - Unique labels [0 1]
2022-07-11 12:33:42.006 | DEBUG    | lapros.noise:__rank_suspicious:27 - Trying to reshape probas
2022-07-11 12:33:42.007 | DEBUG    | lapros.noise:__rank_suspicious:29 - Shape of labels and probas: (5,) vs (5, 1)
2022-07-11 12:33:42.007 | DEBUG    | lapros.noise:__rank_suspicious:37 - ranks [0.5 0.6 0.7 0.2 0.1]





array([0.5, 0.6, 0.7, 0.2, 0.1])

suspect(
    pd.Series([1, 0, 0, 1, 1]),
    probas=pd.DataFrame([0.5, 0.6, 0.7, 0.8, 0.9]),
)

2022-07-11 12:33:42.063 | DEBUG    | lapros.noise:suspect:61 - Pandas series labels and dataframe probas must have same length 5 vs 5
2022-07-11 12:33:42.064 | DEBUG    | lapros.noise:__rank_suspicious:22 - Numpy arrays labels and probas must have same length 5 vs 5
2022-07-11 12:33:42.065 | DEBUG    | lapros.noise:__rank_suspicious:25 - Unique labels [0 1]
2022-07-11 12:33:42.065 | DEBUG    | lapros.noise:__rank_suspicious:29 - Shape of labels and probas: (5,) vs (5, 1)
2022-07-11 12:33:42.065 | DEBUG    | lapros.noise:__rank_suspicious:37 - ranks [0.5 0.6 0.7 0.2 0.1]





0    0.5
1    0.6
2    0.7
3    0.2
4    0.1
dtype: float64