Awesome KAN is a curated collection of resources focused on Kolmogorov-Arnold Networks, organized to serve researchers and developers exploring this neural network architecture.
The repository addresses the need for a centralized reference point in the KAN field by aggregating libraries, projects, tutorials, papers, and other materials. Rather than implementing KAN functionality itself, the project functions as a discovery and navigation tool, organizing resources across multiple categories including foundational papers, theoretical work, implementations, benchmarks, and educational content. This approach helps practitioners quickly locate relevant tools and knowledge without searching across scattered sources.
Developers should adopt this resource if they are entering the KAN field or seeking to understand the landscape of available implementations and research. The collection suits anyone building KAN-based projects who needs to evaluate existing libraries, review benchmark results, or study the underlying theory. The repository's organization into distinct sections—covering papers, libraries, library-based implementations, convolutional variants, benchmarks, non-Python implementations, and tutorials—makes it useful both for initial exploration and for finding specific resource types.
The project maintains an active collection with regular contributions from the community. The repository accepts pull requests to expand its coverage of KAN-related resources. The curation reflects ongoing engagement with the field as new libraries, papers, and projects emerge in the Kolmogorov-Arnold Network space.
