# shenweichen/DeepMatch

A deep matching model library for recommendations & advertising. It's easy to train models and to export representation vectors which can be used for ANN search.

Repository: https://github.com/shenweichen/DeepMatch
Canonical: https://ross.abutalabs.com/products/deepmatch
Homepage: https://deepmatch.readthedocs.io/en/latest/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: dssm, youtubednn, mind, collaborative-filtering, factorization-machines, matching, recommendation, comirec
Last push: 2026-04-18T05:14:22+00:00

## Health v2 (maintenance only)
Score: 72/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 78, release rhythm 48, longevity 100
- inputs: {"age_days": 2340, "days_push": 137, "days_rel": 137, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2433, forks 541 (observed 2026-08-28T04:06:51.245461+00:00)

## What it is
DeepMatch is a Python library of deep matching models for recommendations and advertising, built on TensorFlow/Keras. It lets users train models with fit/predict and export user and item representation vectors for approximate nearest neighbor search.

## Use cases
- train a DSSM two-tower matching model for recommendations
- export user and item embedding vectors for ANN retrieval
- implement YouTubeDNN candidate generation for a recommender
- build deep matching models for online advertising
- experiment with MIND and ComiRec multi-interest models
- train factorization-machine style matching models in Python

## When to choose
- you need ready implementations of deep matching models like DSSM, YouTubeDNN, or MIND
- you want to export embeddings for vector/ANN search in a recommendation pipeline
- you prefer a Keras-style fit/predict workflow on TensorFlow

## When to avoid
- you need a full end-to-end recommender system with serving and ranking, not just matching models
- your stack is PyTorch-based rather than TensorFlow
- you need non-neural or lightweight collaborative filtering only

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, search-engine
- domain: machine-learning, deep-learning
- platform: python
- tags: recommendation-systems, matching-models, dssm, youtubednn, ann-search, tensorflow, advertising, collaborative-filtering, recommendation

## Member repositories
- shenweichen/DeepMatch (main) score 72

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:51.245461+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 2026-08-30T02:31:10.220247+00:00, confidence not recorded.
  - readme: https://github.com/shenweichen/DeepMatch (fetched 2026-08-28T04:06:51.245461+00:00, sha ab25fc3a2e1e)
  - registry_pypi: https://pypi.org/pypi/deepmatch/json (fetched 2026-08-29T10:13:02.738292+00:00, sha 807120747c3d)
- Data as of 2026-08-30T08:39:29.467469+00:00.