XplainML 0.0.1

XplainML is a comprehensive Python package designed for Explainable AI (XAI) and Responsible AI practices. It provides a suite of tools and algorithms to enhance the transparency, interpretability, and fairness of machine learning models.

XplainML: Empowering Transparent and Responsible AI

Welcome to XplainML – your ultimate toolkit for unlocking the power of transparent and responsible AI! Designed and developed by Deependra Verma, XplainML empowers data scientists, machine learning engineers, and AI practitioners to understand, interpret, and trust their models with ease.

Introduction

XplainML is an open-source Python package designed to provide transparent and responsible AI capabilities to users. With XplainML, you can easily interpret your AI models, detect and mitigate bias, ensure fairness, and promote ethical AI practices.

Features

Explainable AI (XAI) Made Easy

Unravel the mysteries of your AI models with intuitive explanations using state-of-the-art techniques like SHAP and LIME.

Responsible AI Integration

Detect and mitigate bias, ensure fairness, and promote ethical AI practices with built-in fairness metrics and bias mitigation algorithms.

Installation

To install XplainML, simply run:

pip install XplainML

Usage

Explainable AI (XAI)

SHAP Explanations

# Import the necessary libraries
import shap
from sklearn.datasets import load_iris
from sklearn.ensemble import RandomForestClassifier

# Load the Iris dataset
iris = load_iris()
X, y = iris.data, iris.target

# Train a random forest classifier
model = RandomForestClassifier(n_estimators=100, random_state=42)
model.fit(X, y)

# Initialize the SHAP explainer
explainer = shap.Explainer(model)

# Generate SHAP explanations for a sample instance
shap_values = explainer(X[:1])

# Visualize the SHAP explanations
shap.plots.waterfall(shap_values[0])

LIME Explanations

# Import the necessary libraries
import lime
import lime.lime_tabular
from sklearn.datasets import load_iris
from sklearn.ensemble import RandomForestClassifier

# Load the Iris dataset
iris = load_iris()
X, y = iris.data, iris.target

# Train a random forest classifier
model = RandomForestClassifier(n_estimators=100, random_state=42)
model.fit(X, y)

# Initialize the LIME explainer
explainer = lime.lime_tabular.LimeTabularExplainer(X, feature_names=iris.feature_names, class_names=iris.target_names)

# Generate LIME explanations for a sample instance
explanation = explainer.explain_instance(X[0], model.predict_proba)

# Visualize the LIME explanations
explanation.show_in_notebook()

Responsible AI Integration

Bias Detection

# Import the necessary libraries
from aif360.datasets import GermanDataset
from aif360.metrics import BinaryLabelDatasetMetric

# Load the German credit dataset
dataset = GermanDataset()
privileged_group = [{'sex': 1}]
unprivileged_group = [{'sex': 0}]

# Compute metrics for bias detection
metric = BinaryLabelDatasetMetric(dataset, unprivileged_group=unprivileged_group, privileged_group=privileged_group)
print("Mean Difference: ", metric.mean_difference())
print("Disparate Impact: ", metric.disparate_impact())

Bias Mitigation

# Import the necessary libraries
from aif360.algorithms.preprocessing import Reweighing

# Apply bias mitigation using Reweighing
biased_dataset = dataset.convert_to_dataframe()[0]
rw = Reweighing(unprivileged_groups=unprivileged_group, privileged_groups=privileged_group)
biased_dataset = rw.fit_transform(biased_dataset)

# Verify bias mitigation results
metric_biased = BinaryLabelDatasetMetric(biased_dataset, unprivileged_group, privileged_group)
print("Mean Difference after mitigation: ", metric_biased.mean_difference())
print("Disparate Impact after mitigation: ", metric_biased.disparate_impact())

Contributing

We welcome contributions from the community! Whether it's fixing bugs, adding new features, or improving documentation, your contributions help make XplainML better for everyone. Check out our Contributing Guidelines to get started.

License

XplainML is licensed under the MIT License. See the LICENSE file for details.

About the Author

Deependra Verma
Data Scientist
Email | LinkedIn | GitHub | Portfolio