ikpykit 0.4.0

About The Project

IKPyKit (Isolation Kernel Python toolKit) is an intuitive Python library designed for a variety of machine learning tasks including kernel similarity calculation, anomaly detection, clustering, and change detection—all powered by the innovative Isolation Kernel (IK) . Isolation Kernel is a data-dependent kernel that measures similarity by isolating data points using an isolation mechanism. It uniquely adapts to the data distribution, with the property that points in sparse regions are more similar than those in dense regions. Notably, it requires no learning or closed-form expression, making it efficient and scalable.

Why use Isolation Kernel?

• Data-Dependent Similarity: Unlike traditional kernels (e.g., Gaussian, Laplacian), Isolation Kernel adapts to the structure of the data rather than assuming a fixed similarity function.
• Handles Sparse and Dense Regions: Isolation Kernel effectively accounts for varying data densities, making it ideal for datasets with non-uniform distributions.
• No Learning Required: It eliminates the need for training or parameter tuning, simplifying implementation while reducing computational cost.
• Effective in High Dimensions: It uniquely addresses the curse of dimensionality, being the only known measure capable of finding exact nearest neighbors in high-dimensional spaces.
• Versatile Applications: Isolation Kernel has been successfully applied to tasks like anomaly detection, clustering, and processing stream data, graph data, trajectory data, and more.

Learn more about its history and development on the IsolationKernel GitHub page.

Why use IKPyKit?

IKPyKit is specifically built to harness the power of Isolation Kernel, providing specialized algorithms for a wide range of data types and tasks. Its seamless integration with the scikit-learn API allows easy adoption and compatibility with scikit-learn tools.

• Tailored for Isolation Kernel: IKPyKit directly leverages the unique properties of Isolation Kernel for efficient and effective machine learning solutions.
• Efficient and User-Friendly: Designed for simplicity and performance, IKPyKit offers an intuitive interface built on the scikit-learn API.
• Support for Diverse Data Types: It supports graph data, group data, stream data, time series, and trajectory data, making it versatile for various domains.
• Comprehensive Resources: Users benefit from rich documentation and examples to quickly understand and apply the library’s features.
• Ideal for Research and Industry: IKPyKit is suitable for both academic research and industrial applications, providing scalable and cutting-edge tools for modern machine learning challenges.

Installation & Dependencies

Recommended installation (with uv):

Install uv first: https://docs.astral.sh/uv/getting-started/

uv pip install ikpykit

If you prefer classic pip:

pip install ikpykit

For more installation options, including dependencies and additional features, check out our Installation Guide.

Example

# Anomaly Detection using inne.
import numpy as np
from ikpykit.anomaly import INNE
X = np.array([[-1.1, 0.2], [0.3, 0.5], [0.5, 1.1], [100, 90]])
clf = INNE(contamination=0.25).fit(X)
clf.predict([[0.1, 0.3], [0, 0.7], [90, 85]])

Implemented Algorithms

Summary

(i) Isolation Kernel:

(ii) Point Anomaly Detection:

(iii) Point Clustering:

(iv) Graph Mining:

(v) Group Mining:

(vi) Stream Mining:

(vii) Trajectory Mining:

(viii) Time Series Mining:

Features

IKPyKit provides a set of key features designed to make machine learning tasks easy and efficient. For a detailed overview, see the User Guides.

Examples and tutorials

Explore our extensive list of examples and tutorials to get you started with IKPyKit. You can find them here.

How to contribute

Primarily, IKPyKit development consists of adding and creating new algorithms, new validation strategies, or improving the performance of the current code. However, there are many other ways to contribute:

• Submit a bug report or feature request on GitHub Issues.
• Contribute a Jupyter notebook to our examples.
• Write unit or integration tests for our project.
• Answer questions on our issues, Stack Overflow, and elsewhere.
• Translate our documentation into another language.
• Write a blog post, tweet, or share our project with others.

For more information on how to contribute to IKPyKit, see our Contribution Guide.

Visit our authors section to meet all the contributors to IKPyKit.

Citation

If you use IKPyKit for a scientific Publication, we would appreciate citations to the Publication software.

BibTeX:

@software{IKPyKit,
    author = {Xin Han, Yixiao Ma, Ye Zhu, and Kaiming Ting},
    title = {IKPyKit：A Python Library for Isolation Kernel Toolkit},
    version = {0.1.0},
    month = {3},
    year = {2025},
    license = {BSD-3-Clause},
    url = {https://github.com/IsolationKernel/ikpykit}
}

License

BSD-3-Clause License