IBM AI Fairness 360
Pricing Details
Free, open-source software under the Apache 2.0 license.Features
Bias Detection Metrics, Bias Mitigation Algorithms (Pre-processing, In-processing, Post-processing), Extensible Framework, Integration with ML Frameworks (TensorFlow, PyTorch, Scikit-learn), Documentation, Tutorials.Integrations
Integrates into Python-based ML workflows, compatible with TensorFlow, PyTorch, Scikit-learn.Preview
IBM AI Fairness 360 (AIF360) is an open-source toolkit developed by IBM Research to help developers, data scientists, and researchers detect, understand, and mitigate bias in machine learning (ML) models throughout the artificial intelligence (AI) application lifecycle. Released in 2018 under the Apache 2.0 license, AIF360 provides an extensive library of metrics (over 70) for quantifying various types of unfairness and bias in datasets and ML models. Additionally, it includes a set of over 10 state-of-the-art algorithms for bias mitigation that can be applied at different stages of the ML workflow: pre-processing the data, in-processing during model training, and post-processing after obtaining prediction results. AIF360 is designed as an extensible framework, allowing researchers and practitioners to contribute new metrics and algorithms and benchmark their performance. The toolkit integrates with popular ML libraries in Python, such as TensorFlow, PyTorch, and Scikit-learn. It is not an end-user product but a tool for developers of AI systems who aim to ensure their fairness and ethical behavior. Using AIF360 contributes to increasing trust in AI systems and helps organizations comply with growing demands for algorithmic ethics and transparency. The platform is a Python library compatible with various operating systems that support Python.