This version uses the revised math for SRW (with M&dM, p&dp, and gradient steps).

Training is much faster and duplicates less computation. Inheritance among the SRW classes is cleaner and requires fewer lines of code for overrides. Regularization is slightly simpler.

2.1 runnables are the same as 2.0 and include Grounder, Trainer, QueryAnswerer, and GradientFinder.

This version uses our first draft of the SRW and prover machinery.

SRW includes parallel package structures for storing graphs using Java collections (which are easier to develop because you can see what the nodes are in a debugger) or GNU trove collections (which are faster, but convert all the nodes to ID numbers & is fairly unintelligible in a debugger). SRW vectors all use Java collections.

Provers are based on LogicProgramStates and use a Theta object to track variable assignments.

This version's runnables include ExampleCooker, Trainer, Tester, QueryAnswerer, GradientFinder, and Experiment.

Notable shortcomings of this version are minAlpha exceptions and issues resolving theta.

This version uses an updated prover model based on WAM, the Warren Abstract Machine, and nominal changes to the SRW system.

The 2.0 prover is much faster and has a more sensical logic program plugin architecture.

The 2.0 SRW is no longer duplicated between Java collections and trove, but uses a unified graph storage system that is more compact than a pure-Java solution yet more readable than pure-trove.

2.0 runnables include Grounder, Trainer, QueryAnswerer, and GradientFinder.