OSLa (On-line Structure Learning)

On-line Structure Learning (OSL and OSLa) (lomrf slearn) is an extension of LoMRF, an open-source implementation of Markov Logic Networks.

License

Copyright (c) 2014 – 2018 Anastasios Skarlatidis and Evangelos Michelioudakis
LoMRF is licensed under the Apache License, Version 2.0: https://www.apache.org/licenses/LICENSE-2.0

Features

  1. Parallel grounding algorithm based on Akka Actors library.
  2. Marginal (MC-SAT) and MAP (MaxWalkSAT and LP-relaxed Integer Linear Programming) inference (lomrf infer).
  3. Batch and on-line Weight Learning (Max-Margin, AdaGrad and CDA) (lomrf wlearn).
  4. On-line Structure Learning (OSL and OSLa) (lomrf slearn).
  5. MLN knowledge base compilation (lomrf compile):
  • Predicate completion.
  • Clausal form transformation.
  • Replacement of functions with utility predicates and vice versa.
  • Reads and produces Alchemy compatible MLN files.
  1. Can export ground MRF in various formats (lomrf export).
  2. Can compare MLN theories (lomrf diff).
  3. Online supervision completion on semi-supervised training sets [currently experimental] (lomrf supervision)