# KindXiaoming/pykan

Kolmogorov Arnold Networks

Repository: https://github.com/KindXiaoming/pykan
Canonical: https://ross.abutalabs.com/products/pykan
Language: Jupyter Notebook
License: MIT
License Family: permissive
Last push: 2025-01-19T18:49:39+00:00

## Health v2 (maintenance only)
Score: 27/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 2, release rhythm 40, longevity 61
- inputs: {"age_days": 858, "days_push": 591, "days_rel": 657, "gap_med": 30, "n_releases_24m": 2}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 16334, forks 1559 (observed 2026-08-28T04:11:15.102418+00:00)

## What it is
Pykan is a Python library implementing Kolmogorov-Arnold Networks (KANs), an alternative to multi-layer perceptrons with learnable activation functions on edges, built on PyTorch. It emphasizes improved accuracy and interpretability, including symbolic regression and visualization of learned functions.

## Use cases
- train interpretable neural networks instead of MLPs
- perform symbolic regression to discover formulas from data
- fit scientific models like PDEs with learnable activation networks
- visualize and prune neural network architectures
- experiment with Kolmogorov-Arnold representation theorem based models

## When to choose
- you need more interpretable models than standard MLPs
- you want to extract symbolic formulas from learned networks
- your research explores alternatives to conventional neural architectures
- you work on small-to-medium scientific ML problems runnable on CPU

## When to avoid
- you need highly optimized GPU-scale training throughput
- you require a mature ecosystem with broad production tooling
- your workload depends on the standard MLP tooling and hardware acceleration
- you need fast inference at scale, since symbolic branches can be slow

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, data-science, math
- domain: machine-learning, deep-learning, data-science
- platform: python, cross-platform
- tags: kolmogorov-arnold-networks, kan, pytorch, interpretable-ml, neural-networks, symbolic-regression, algorithms

## Member repositories
- KindXiaoming/pykan (main) score 27

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:15.102418+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 2026-08-29T17:05:08.760046+00:00, confidence not recorded.
  - readme: https://github.com/KindXiaoming/pykan (fetched 2026-08-28T04:11:15.102418+00:00, sha 79c78d6486b2)
- Data as of 2026-08-30T08:39:29.467469+00:00.