Skip to content

Latest commit

 

History

History
42 lines (31 loc) · 2.64 KB

CHANGELOG.md

File metadata and controls

42 lines (31 loc) · 2.64 KB
Raw

0.5.3

  • Support for UTF-8-BOM and UTF-16 character encodings
  • Minor bug fixes and improvements

0.5.2

  • Minor bug fixes and improvements. delSAT -> diff-SAT

0.5.1

  • Bug fix: in rare cases, conflict nogood generation messed up the memory layout

0.5.0

  • Optimized SAT solver core (and thus improved sampling performance)
  • User API for solving problems with probabilistic or non-probabilistic CNF clauses and ASP ground rules
  • Direct User API support for probabilistic non-ground normal rules
  • aspif and User API support for weight and choice rules, double default negation in rule bodies and for double and single negation in rule heads
  • Probabilistic ASP Intermediate Format (PASPIF) (ASPIF plus a new probabilistic rule type)
  • CDNL interleaved with Stochastic Local Search (WalkSAT or Simulated Annealing (SASAT-style)) for regular SAT solving
  • Bug fixes and several minor improvements
  • Some code cleanup and refactoring

0.4.1

  • Support for weighted clauses annotated with probabilities in DIMACS CNF (Probabilistic CNF). See Usage for details.
  • _eval_("term", "?") meta-predicate for ad hoc querying of probabilities and other statistics (e.g., remaining loss) over sampled models. An example can be found under Performing ad hoc queries.
  • Argument -s k (k>1) for sampling k models uniformly from the multimodel cost solution.
  • Argument -n -k (k>1) for sampling k models such that k-m of these models are sampled uniformly from the sample (with size m) for which the cost threshold was initially reached.
  • Improved finite differences approach to the case where not all parameter atoms are measured (i.e., occur in costs), automatically activated when needed. Switch useNumFiniteDiff is not required anymore, however, purely numerical differentiation approach can still be enforced with --solverarg mixedScenario false (in rare cases this is more efficient).
  • Typically higher solution entropy for the same sample size (-n), compared to previous versions. For how to increase the entropy further see README.md.
  • Support for non-quoted _cost_ and _eval_ terms using plain logic programming term notation (e.g., _cost_(pow(subtr(div(77,100),f(coin(1,heads))),2)) instead of _cost_("(0.77-f(coin(1,heads)))^2")).
  • Several minor improvements and bug fixes
  • Enhanced documentation (this file)

0.4.0

  • Simplified usage; support for parameter atom and cost function specification using special logical predicates
  • Support for distinct sets of parameter atoms and measured atoms, enabling preliminary support for inductive/abductive inference.