dvdmjohnson / d3m_michigan_primitives

Contains primitives developed by the University of Michigan team as part of the Data Driven Discovery (D3M) project
Other
2 stars 0 forks source link
d3m

spider -- The University of Michigan DARPA D3M SPIDER Project CodeBase

This codebase includes all methods and primitives that comprise the SPIDER project. These do include the featurization primitives that we contributed as well as the featurization baseclass that provides a standard protocol for accessing and implementing the featurization across the D3M program.

Maintained by Jason Corso, Laura Balzano and DARPA D3M SPIDER project team members at the University of Michigan.

Brief Description and File System Layout

spider/                     --->  main package
  cluster/                  --->  sub-package for clustering primitives
  pipelines/                --->  contains python code defining example pipelines using Michigan primitives; can be used to generate "runnable" JSON pipelines
  supervised_learning/      --->  sub-package for supervised learning primitives and examples
  tests/                    --->  sub-package with unit tests 
  unsupervised_learning/    --->  sub-package for learning linear subspace from unlabelled data

Primitives Included

# Cluster Primitives
spider.cluster.ekss
spider.cluster.kss
spider.cluster.ssc_cvx
spider.cluster.ssc_admm
spider.cluster.ssc_omp

# Supervised Learning Primitives
spider.supervised_learning.owl

# Unsupervised Learning Primitives
spider.unsupervised_learning.grasta

License

MIT license.