MADlib Roadmap

Jump to bottom Edit New page

cwelton edited this page Aug 14, 2012 · 8 revisions

MADlib Development timeline

MADlib is currently on a quarterly release cadence with monthly development sprints.

MADlib v0.5

June 14th: Call for features
- Multinomial Logistic Regression (MADLIB-575)
- Cox Proportional Hazard Models (MADLIB-576)
- Marginal effects for ols, logistic, and multivariate logistic
- Development of general convex optimization architecture, including support for:
  - Neural Networks (MADLIB-424)
  - LASSO & Ridge Regression (MADLIB-418)
  - Performance and scalability for SVM (MADLIB-627)
  - Conditional Random Fields (MADLIB-628)
- Performance and scalability improvements
  - K-means (MADLIB-522,MADLIB-454)
  - Decision Trees and Random Forest (MADLIB-445,MADLIB-483,MADLIB-496)
  - Association Rules (MADLIB-580)
- Miscellaneous Usability features
  - Improved reporting of output for Association Rules (MADLIB-411)
- Improvements to testing framework
July 26th: Design Specifications finalized for new content
Aug 2nd: Status Checkpoint, push content that does not look on track for dev complete to next release.
Aug 20th: Dev Complete
Sep 27th: Testing certification complete
Oct 4th: Estimated Ship date MADlib v0.5

MADlib v1.0

Sep 11th: Call for features
- Items deferred from v0.5
  - Kalman Filters (MADLIB-423)
- Usability and consistency of interface
  - More consistent user API across modules (MADLIB-587)
- Improvements to testing framework
Oct 11th: Rough timeline for the release established