Candidate Some potential roadmap for remainder of 2016:ideas for 2018+ below.  For more details please see the JIRAs. 

Predictive models

...

• Multi-class SVM 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1037

• Mixed effects modeling 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-987

• Gradient boosted machines
• Gaussian Mixture Model using Expectation Maximization (EM)

...

• algorithm 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-410

• Hierarchical clustering
• k-NN improvements 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1061

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1181

• Deep learning

Graph

• Personalized PageRank 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1084

• Betweenness centrality 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1121

• Graph cut 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1074

• Triangle counting 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1125

• Minimum spanning tree
• Eigenvector centrality 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1123

• APSP performance improvements 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1155

Utilities

• Balanced datasets 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1168

  
  Summary - add more statistics 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1167

• Anonymization  

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-911

• URI tools  

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-910

Graph

Shortest path  MADLIB-992
Standard traversal
  depth first search
  breadth first search
  topological sort
One mode projection (converting a bi-partitite graph of user-item graph to user-user or item-item graph)
Connected components
Page rank
Hierarchical graph cut
Between-ness centrality
Minimum spanning tree

Utilities

...

Usability

• Expand coverage for PivotalR

...

• Interface improvement and consistency

...

• for 2.0 release (does not need to be backward compatible) 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1157

• Python API

Performance and scalability

Work around PostgreSQL 1 GB field size limit  MADLIB-991

Platform

• Mini-batching 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1048

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1037

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1200

• Improve decision tree and random forest performance for run-time and memory usage 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1057

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-976

Platforms and Frameworks

• PostgreSQL 10 support 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1185

• Support modern versions of gcc 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key MADLIB-1025

• Tensorflow support, or another deep learning framework
• GPU support

...