Date post: | 15-Apr-2017 |
Category: |
Software |
Upload: | jo-fai-chow |
View: | 1,930 times |
Download: | 0 times |
Brendan Herger [email protected]
Why are we here?There are many awesome,
competing, overlapping machine learning libraries
And people usually stick to one
R / CRANStrengths
• Algorithm selection - CRAN Library
• Community
• Commercial backing - Revolution Analytics / RStudio
• Packages (Plotting - GGPlot 2, Time series - SARIMA / GARCH)
R / CRANWeaknesses
• Algorithm implementation
• Documentation
• Computational limits
• Ongoing support
Pandas / SKLearnStrengths
• Unified APIs
• Algorithm Choice / quick iteration
• Documentation
• Other Python packages
Pandas / SKLearnWeaknesses
• Limited to one machine
• Two packages shimmed together
• Python broilerplate
Pandas / SKLearnRoadmap
• Python: Multithreading (Project Ibis)
• Pandas: Rapid development, better time series, better SKLearn integration
• SKLearn: C Implementations of some algorithms, awaiting new algorithms
Thanks! [email protected]