kkroy36/rfgb

Developing gradient boosting for relational data

Relational Functional Gradient Boosting (RFGB)

RFGB in Python. Inspired by BoostSRL.

Getting Started

Prerequisites

• Python (2.7, 3.3, 3.4, 3.5, 3.6)

Installation

None currently. Clone the repository with git or download a zip archive, then run the scripts from the command line.

Sample Usage

1. Logistics

    python src/main.py -target unload -train testDomains/Logistics/train/ -test testDomains/Logistics/test/ -trees 10
   

Classification with Expert Advice (-expAdvice)

Preferred and non-preferred labels may be provided as advice during classification via logical rules. This advice may be specified in a file named advice.txt in the train directory for a dataset.

Four datasets (BlocksWorld, HeartAttack, Logistics, and MoodDisorder) have an advice file included.

1. Logistics

    python src/main.py -expAdvice -target unload -train testDomains/Logistics/train/ -test testDomains/Logistics/test/ -trees 10
   

2. HeartAttack

    python src/main.py -expAdvice -target ha -train testDomains/HeartAttack/train/ -test testDomains/HeartAttack/test/ -trees 10
   

Targets

"Targets" specify what is learned, examples of the target are provided in pos.txt, neg.txt, or examples.txt (for regression). These are specified here for convenience.

Dataset	Target
BlocksWorld	putdown
BostonHousing	medv
HeartAttack	ha
Insurance	value
Logistics	unload
MoodDisorder	bipolar
TicTacToe	put or dontput
ToyCancer	cancer
XOR	xor

In Development

• Test cases (codecov >90%)
• Learning Markov Logic Networks
• Learning with Soft-Margin

License

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

A full copy of the license is available in the base of this repository. For more information, see https://www.gnu.org/licenses/

Acknowledgements

The authors would like to thank Professor Sriraam Natarajan, Professor Gautam Kunapuli, and fellow members of the StARLinG Lab at the University of Texas at Dallas.

rfgb's Issues

License?

Having a license is pretty important for open source software, and also spells out the limitations for things copyright, patents, and source code distribution!

https://choosealicense.com/

Is this still being developed?

Haven't seen major updates since last December. If I'm correct, it would likely be easier to split my fork into a separate project going forward.