# tensorflow/neural-structured-learning

Training neural models with structured signals.

Repository: https://github.com/tensorflow/neural-structured-learning
Canonical: https://ross.abutalabs.com/products/neural-structured-learning
Homepage: https://www.tensorflow.org/neural_structured_learning
Language: Python
License: Apache-2.0
License Family: permissive
Topics: tensorflow, graph-learning, adversarial-learning, neural-networks, regularization, structured-signals, keras
Last push: 2026-07-21T20:11:25+00:00

## Health v2 (maintenance only)
Score: 65/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 93, release rhythm 8, longevity 100
- inputs: {"age_days": 2563, "days_push": 43, "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 1010, forks 189 (observed 2026-08-28T04:03:13.040155+00:00)

## What it is
Neural Structured Learning (NSL) is a TensorFlow framework for training neural networks with structured signals, either explicit graphs or implicit adversarial perturbations. It provides Keras and Estimator APIs plus tools for building graphs, improving accuracy with limited labels and robustness against adversarial attacks.

## Use cases
- train neural networks with graph-structured data
- improve model accuracy when labeled data is scarce
- make models robust against adversarial attacks
- semi-supervised learning with unlabeled samples
- add adversarial perturbation regularization to Keras models
- build graphs from training data for neural graph learning

## When to choose
- you already use TensorFlow/Keras and have relational or graph structure among samples
- you need adversarial training for model robustness
- you have limited labeled data plus unlabeled samples with similarity signals

## When to avoid
- your stack is PyTorch or JAX
- you need the latest TensorFlow 2.x features or active development
- your data has no relational structure and you don't need adversarial robustness

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, llm-training
- domain: machine-learning, deep-learning, artificial-intelligence, security
- platform: python, cross-platform
- tags: tensorflow, keras, graph-learning, adversarial-learning, regularization, semi-supervised-learning, neural-graph-learning

## Member repositories
- tensorflow/neural-structured-learning (main) score 65

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:13.040155+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 2026-08-30T07:11:58.778517+00:00, confidence not recorded.
  - readme: https://github.com/tensorflow/neural-structured-learning (fetched 2026-08-28T04:03:13.040155+00:00, sha db500fc31441)
  - homepage: https://www.tensorflow.org/neural_structured_learning (fetched 2026-08-29T13:12:16.675046+00:00, sha 70edada274fa)
  - site_page: https://www.tensorflow.org/install (fetched 2026-08-29T13:12:16.683835+00:00, sha 584a762da891)
  - site_page: https://www.tensorflow.org/tfx/api_docs (fetched 2026-08-29T13:12:16.687572+00:00, sha 6977825696fe)
  - site_page: https://www.tensorflow.org/about (fetched 2026-08-29T13:12:16.690807+00:00, sha 817250744d91)
  - site_page: https://www.tensorflow.org/about/case-studies (fetched 2026-08-29T13:12:16.692459+00:00, sha e6701029eec0)
  - site_page: https://www.tensorflow.org/neural_structured_learning/api_docs/python/nsl (fetched 2026-08-29T13:12:16.694410+00:00, sha 021be442b3fb)
  - site_page: https://www.tensorflow.org/neural_structured_learning/install (fetched 2026-08-29T13:12:16.696165+00:00, sha 5544fe10c9a6)
  - registry_pypi: https://pypi.org/pypi/neural-structured-learning/json (fetched 2026-08-29T13:12:16.697865+00:00, sha 0bd60ada958b)
  - site_page: https://www.tensorflow.org/community/contribute (fetched 2026-08-29T13:12:16.689241+00:00, sha 77d394fa76e7)
- Data as of 2026-08-30T08:39:29.467469+00:00.