# Blealtan/efficient-kan

An efficient pure-PyTorch implementation of Kolmogorov-Arnold Network (KAN).

Repository: https://github.com/Blealtan/efficient-kan
Canonical: https://ross.abutalabs.com/products/efficient-kan
Language: Python
License: MIT
License Family: permissive
Last push: 2024-08-01T12:21:05+00:00

## Health v2 (maintenance only)
Score: 24/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 60
- inputs: {"age_days": 853, "days_push": 762, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4655, forks 414 (observed 2026-08-28T04:08:55.691145+00:00)

## What it is
An efficient pure-PyTorch implementation of Kolmogorov-Arnold Networks (KAN) that reformulates the computation as matrix multiplications over B-spline basis functions. It significantly reduces memory cost compared to the original implementation while supporting forward and backward passes.

## Use cases
- train Kolmogorov-Arnold Networks efficiently in PyTorch
- replace MLP layers with KAN layers in experiments
- run KAN models on MNIST-style classification tasks
- experiment with spline-based learnable activation functions
- reduce memory usage when training KAN models

## When to choose
- you want a fast, memory-efficient KAN implementation in pure PyTorch
- you need KAN layers that integrate with standard PyTorch training loops
- the original pykan implementation is too slow or memory-hungry for your workload

## When to avoid
- you need the exact L1-on-activations regularization from the original KAN paper
- you require the full feature set and interpretability tooling of the original pykan
- you need a framework other than PyTorch

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning
- domain: machine-learning, deep-learning
- platform: python
- tags: kolmogorov-arnold-network, kan, pytorch, neural-networks, b-splines

## Member repositories
- Blealtan/efficient-kan (main) score 24

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:55.691145+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:19:28.582097+00:00, confidence not recorded.
  - readme: https://github.com/Blealtan/efficient-kan (fetched 2026-08-28T04:08:55.691145+00:00, sha 83ea0bfb7103)
- Data as of 2026-08-30T08:39:29.467469+00:00.