NeuroB

An artificial deep learning framework for ProB.

With NeuroB, one can set up training data for machine learning problems related to the B method. Using ProB2 to load B machines and to access their state spaces, one can create training sets from B predicates or whole B machines and set up classification or regression tasks over them.

Requirements

• Java 1.8+
• Z3 Theorem Prover, version 4.8.9

Getting the development started

All dependencies reside in the build.gradle file. For IntelliJ, one can simply import it:

idea build.gradle

Modules contained in the repository

• core: NeuroB library itself and all its code,
• cli: Command line program to use NeuroB from command line.

Coding conventions

• indentation
  • 4 spaces
  • hanging indent: 4 spaces as well
• maximal line length: 100 characters
• no wildcard imports
• src/ layout per module:
  • src/main/java: java source code
  • src/main/resources: non-java resource files
  • src/test/java: unit tests
  • src/it/java: integration tests
  • src/it/resources: integration test resources (B machines and alike)