Skip to content

Latest commit

 

History

History
38 lines (31 loc) · 1.94 KB

README.md

File metadata and controls

38 lines (31 loc) · 1.94 KB

Learning Partial Boolean Functions by Local Search

PROOF (for PaRtial bOOlean Functions) is my implementation for my research project at the chair of Machine Learning for Computer Vision at Technische Universität Dresden.

Dependencies

  • conda for preparing the data
  • rust to compile PROOF

Usage

DATA

NOTE: The data format will change soon and might change again at any time if that makes accessing the data easier. This, combined with the wish to keep this repository small, lead to the decision that only the raw data and a conversion script will be provided. To run PROOF the data must be recreated first!

For details on the chosen data format and how to recreate the data see here.

Compilation

Simply run cargo build --release to compile PROOF. The --release here makes for a huge improvement in runtime that is well worth the extra seconds in compile time. Optionally, you would want to symlink the resulting binary for easy access by ln -s target/release/proof proof

Running PROOF

PROOF can be configured via proof -c [config_file]. For testing purposes the testing_config.json is provided.

NOTE: It is highly recommended to only run proof with the testing configuration for now, as its runtime is still to slow to tackle real world circumstances.

Verbosity of debug information can be increased by each additional -v.

Licensing

This project is licensed under the GNU General Public License v3 except for those parts (lines of code from libraries used in this project) already licensed under other licenses.

Copyright

This projects structure is based on a custom version of Georg Lauterbach's themis template for command line rust programs.