There are huge amounts of repos on Github. People tend to look for high star repos, which are most likely to be high quality. However, the long-tail law also applies to Github, meaning there exist many low star but high code quality repos.

I've collected the so called low-star-high-quality repos list. So far all of them are about Deep Learning and CTR algorithms.

The list is still growing now, and if you like to name some of your findings, any recommendation of any technology is welcome.

The current definition of

• low star: <= 50 stars
• high quality: the code is readable, runnable and highly-efficient is most appreciate

Word2vec:

• https://github.com/lomyal/simple-word-embedding

• https://github.com/akb89/word2vec

• https://github.com/DerekChia/word2vec_numpy

Wide&Deep:

• https://github.com/searchlink/wide_deep_model

ESMM:

• https://github.com/wutenghu/esearch-ctcvr-prediction

• https://github.com/R-Stalker/deep_learning_estimator_labels

DIN:

• https://github.com/baymaxKevin/rec-sys

DCN:

• https://github.com/Leon0427/DCN

DeepFM:

• https://github.com/r-terada/deepFM

GES:

• https://github.com/39239580/dev_ops_recom_sys_ser/tree/9027d40223b5d6537fb16625c15db0812c90d33c/code_tools/embeding_util

CTR:

• https://github.com/Hirosora/LightCTR

• https://github.com/DSXiangLi/CTR