Chaining generators through pipelines¶

This tutorial shows how to chain different generators using the provided Pipeline class.

Beware, this pipeline class is not compatible with the sklearn Pipeline class

In [1]:

import badgers
badgers.__version__

import badgers badgers.__version__

Out[1]:

'0.0.14'

In [2]:

import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns

from sklearn.datasets import make_blobs

from badgers.core.pipeline import Pipeline
from badgers.generators.tabular_data.imbalance import RandomSamplingClassesGenerator
from badgers.generators.tabular_data.noise import GaussianNoiseGenerator

import matplotlib.pyplot as plt import numpy as np import seaborn as sns from sklearn.datasets import make_blobs from badgers.core.pipeline import Pipeline from badgers.generators.tabular_data.imbalance import RandomSamplingClassesGenerator from badgers.generators.tabular_data.noise import GaussianNoiseGenerator

Setup random generator¶

In [3]:

from numpy.random import default_rng
seed = 0
rng = default_rng(seed)

from numpy.random import default_rng seed = 0 rng = default_rng(seed)

Load data¶

Here we simply generate fake data using make_blobs from scikit-learn (see documentation)

In [4]:

X, y = make_blobs(centers=4, random_state=0, cluster_std=0.6)

X, y = make_blobs(centers=4, random_state=0, cluster_std=0.6)

Chaining the generators¶

In [5]:

generators = {
    'imbalance': RandomSamplingClassesGenerator(random_generator=rng),
    'noise': GaussianNoiseGenerator(random_generator=rng)
}

pipeline = Pipeline(generators=generators)

params = {
    'imbalance': {'proportion_classes':{0:0.6, 1:0.25, 2:0.1, 3:0.05}},
    'noise': {'noise_std': 0.5}
}



Xt, yt = pipeline.generate(X=X.copy(), y=y, params=params)

generators = { 'imbalance': RandomSamplingClassesGenerator(random_generator=rng), 'noise': GaussianNoiseGenerator(random_generator=rng) } pipeline = Pipeline(generators=generators) params = { 'imbalance': {'proportion_classes':{0:0.6, 1:0.25, 2:0.1, 3:0.05}}, 'noise': {'noise_std': 0.5} } Xt, yt = pipeline.generate(X=X.copy(), y=y, params=params)

Visualization¶

Here is some code to visualize both the original and the transformed data

In [6]:

Xt[:5]

Xt[:5]

Out[6]:

array([[ 1.20572833,  3.45341106],
       [ 1.24022563,  5.06459723],
       [-0.23972431,  3.45272907],
       [ 2.56611455,  4.02556266],
       [-0.18904398,  3.97085931]])

In [7]:

fig, axes = plt.subplots(1, 2, sharex=True, sharey=True, figsize=(8,4))
sns.scatterplot(x=X[:, 0], y=X[:, 1], hue=y, palette=sns.color_palette("tab10")[:4], ax=axes[0])
sns.scatterplot(x=Xt[:, 0], y=Xt[:, 1], hue=yt, palette=sns.color_palette("tab10")[:4], ax=axes[1])

axes[0].set_title('Original')
axes[1].set_title('Transformed')

plt.tight_layout();

fig, axes = plt.subplots(1, 2, sharex=True, sharey=True, figsize=(8,4)) sns.scatterplot(x=X[:, 0], y=X[:, 1], hue=y, palette=sns.color_palette("tab10")[:4], ax=axes[0]) sns.scatterplot(x=Xt[:, 0], y=Xt[:, 1], hue=yt, palette=sns.color_palette("tab10")[:4], ax=axes[1]) axes[0].set_title('Original') axes[1].set_title('Transformed') plt.tight_layout();

Visualizing the number of instances for all classs¶

In [8]:

fig, axes = plt.subplots(1, 2, sharex=True, sharey=True, figsize=(8,4))

classes, nb = np.unique(y, return_counts=True)
axes[0].bar(classes, nb, color=[f'C{i}' for i in range(len(classes))])
axes[0].set_xlabel('classes')
axes[0].set_ylabel('number of instances')

classes_t, nb_t = np.unique(yt, return_counts=True)
axes[1].bar(classes_t, nb_t, color=[f'C{i}' for i in range(len(classes_t))])
axes[1].set_xlabel('classes')
axes[1].set_ylabel('number of instances')

axes[0].set_title('Original')
axes[1].set_title('Transformed')
plt.tight_layout();

fig, axes = plt.subplots(1, 2, sharex=True, sharey=True, figsize=(8,4)) classes, nb = np.unique(y, return_counts=True) axes[0].bar(classes, nb, color=[f'C{i}' for i in range(len(classes))]) axes[0].set_xlabel('classes') axes[0].set_ylabel('number of instances') classes_t, nb_t = np.unique(yt, return_counts=True) axes[1].bar(classes_t, nb_t, color=[f'C{i}' for i in range(len(classes_t))]) axes[1].set_xlabel('classes') axes[1].set_ylabel('number of instances') axes[0].set_title('Original') axes[1].set_title('Transformed') plt.tight_layout();