FastLAS

Fast Learning from Answer Sets

FastLAS is a new system for Inductive Logic Programming. It is based on the Context-dependent Learning from Answer Sets framework used by ILASP. The first release, as described in the paper “Scalable Inductive Logic Programming Incorporating Domain-specific Optimisation Criteria”, presented at AAAI 2020, has several restrictions, which mean that it is not nearly as general as ILASP, but it is significantly more scalable, and allows a user to express their own optimisation criteria.

The second release, as described in the paper “Scalable Non-observational Predicate Learning in ASP”, presented at IJCAI 2021, lifts some of FastLAS1’s restrictsions, allowing Non-observational predicate learning, and learning in the context of background knowledge with multiple answer sets.

The latest version, as described in the paper “Search Space Expansion for Efficient Incremental Inductive Logic Programming from Streamed Data”, allows FastLAS to learn from examples coming from a data stream comprised of windows of examples. The IncrementalLAS algorithm (incorporated into the main FastLAS system) expands the hypothesis space as each window comes in, without having to redo the work for the previous example windows. Crucially, it still maintains FastLAS’s guarantee that the expanded hypothesis space is OPT-sufficient (over the full set of example windows), meaning that the solution returned by IncrementalLAS is guaranteed to be optimal.

Table of Contents

  1. Downloading, Installing and Running FastLAS
  2. Learning Tasks from the AAAI 2020 Paper
    1. Sentence Chunking
    2. CAVIAR
    3. Policy Learning
  3. FastLOAS

For the document with the proofs omitted from the AAAI 2020 paper, please see https://github.com/spike-imperial/FastLAS/blob/master/proofs_document.pdf.

For the document with the proofs omitted from the IJCAI 2021 paper, please see https://github.com/spike-imperial/FastLAS/blob/master/fast_non_opl_proofs.pdf.

For the document with the proofs omitted from the IJCAI 2022 paper, please see https://github.com/spike-imperial/FastLAS/blob/master/incremental_las_proofs.pdf.

Any questions about the FastLAS system should be directed to Mark Law (mark.law09@imperial.ac.uk).