State of Open Source Machine Learning
Vadim Markovtsev
source{d}
64 open ML projects
Alignment
Projects by year
Projects by interface language
Projects by implementation language
Projects by license
☆ Stars
| peak 1 |
< 1000 |
28% (18) |
| peak 2 |
2000 - 4000 |
25% (16) |
| top |
100,000 |
Tensorflow |
| 2nd |
29,000 |
Keras |
| 3rd |
28,000 |
sklearn |
Tensorflow
- Deep learning
- Disclosed in 2015
- Governed by Google
- Written in C++
- Main API is Python
- 1,500 contributors
- 11,500 GitHub issues
Keras
- Deep learning
- Appeared in 2015
- Governed by Francois Chollet
- Written in Python
- API is Python
- 660 contributors
- 7200 GitHub issues
sklearn
- Generic Machine Learning
- Appeared in 2010
- Independent group governance
- Written in Python
- API is Python
- 1080 contributors
- 5000 GitHub issues
Theano - RIP since September 2017
Contributors
Bus factor
How many core developers can be excluded to kill the project?
Wikipedia
Bus factor
Contribution guidelines
Document which describes how to help the project.
Reference.
Contribution guidelines
Code of conduct
Document which specifies the communication norms. It improves
the diversity and helps to maintain a welcoming environment.
Example.
Code of conduct
Open source machine learning
- Half is owned by companies
- Hype started in 2013, peak in 2015
- Python rules
- Apache license is the most popular
- Top: Tensorflow, Keras, sklearn
Open source machine learning
- Bus factor should be >1
- Contribution guidelines
- Code of conduct