Created
May 18, 2022 12:10
-
-
Save MovsisyanM/618a3dcabfb8b3c7af9a9f4f5e8b93eb to your computer and use it in GitHub Desktop.
Balanced generator for 0, 1 binary classification problems
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
def equigen( | |
x: pd.DataFrame, | |
y: pd.Series, | |
batch_size: int = 256, | |
seed: int = seed, | |
preproc: callable = None, | |
preproc_kwargs: dict = {}): | |
"""Balanced generator for 0, 1 binary classification problems""" | |
np.random.seed(seed) | |
# Negative observations | |
x_n = x[np.array(y == 0)].copy() | |
x_n["_label"] = 0 | |
# Positive observations | |
x_p = x[np.array(y) == 1].copy() | |
x_p["_label"] = 1 | |
while True: | |
sample = pd.concat([ | |
x_n.sample(int(batch_size/2), random_state=seed), | |
x_p.sample(int(batch_size/2), random_state=seed) | |
]).sample(frac=1, random_state=seed) | |
sample, labels = preproc(sample, **preproc_kwargs) if preproc else (sample, labels) | |
yield sample, labels |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment