# raminmh/CfC

Closed-form Continuous-time Neural Networks

Repository: https://github.com/raminmh/CfC
Canonical: https://ross.abutalabs.com/products/cfc
Language: Python
License: Apache-2.0
License Family: permissive
Topics: deep-learning, neural-ode, pytorch, recurrent-neural-networks, sequence-models, tensorflow
Last push: 2024-07-05T00:34:15+00:00

## Health v2 (maintenance only)
Score: 23/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 100
- inputs: {"age_days": 2062, "days_push": 790, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1053, forks 164 (observed 2026-08-28T04:03:23.573223+00:00)

## What it is
Reference implementations of Closed-form Continuous-time (CfC) neural networks, a fast closed-form approximation of liquid time-constant networks for sequential and irregularly-sampled data. Includes TensorFlow and PyTorch modules plus training scripts for Physionet, IMDB, XOR, and Walker2d benchmarks.

## Use cases
- model irregularly-sampled time series like clinical ICU data
- train continuous-time recurrent networks for sequence classification
- benchmark CfC against LSTM and ODE-based neural networks
- run reinforcement learning experiments with liquid neural networks
- reproduce results from the Nature Machine Intelligence CfC paper

## When to choose
- you need efficient continuous-time sequence models without ODE solvers
- your data has irregular time sampling
- you want liquid neural network behavior with fast closed-form dynamics

## When to avoid
- you need a maintained production library - use the ncps package instead
- you need standard feedforward or transformer architectures
- you work outside Python/TensorFlow/PyTorch ecosystems

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning
- domain: machine-learning, deep-learning, artificial-intelligence
- platform: python
- tags: liquid-neural-networks, neural-ode, recurrent-neural-networks, sequence-models, continuous-time, pytorch, tensorflow, research-code

## Member repositories
- raminmh/CfC (main) score 23

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:23.573223+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 2026-08-30T06:59:23.261212+00:00, confidence not recorded.
  - readme: https://github.com/raminmh/CfC (fetched 2026-08-28T04:03:23.573223+00:00, sha 1e423e7d6526)
- Data as of 2026-08-30T08:39:29.467469+00:00.