CHANGES

This change log only contains significant changes and/or new features.
Intermediate bugfixes and speed improvements are not listed.

2021/07/19

    ProbLog 2.2 released. Additional new features include:
        - improvements to LFI (speed, bugfixes)
        - more tests and documentation
        - collect, aggregate and scope libraries (cf. docs/source/prolog.rst)
        - Caching of intermediate results using 'ddnnf' (since 2.1.0.40)
        - Many more...

...

2019/03/28

    Dropped support for Python 2.7+

    (NEW) SDDExplicit
        - A new encoding for SDDs equivalent to the encoding used in d-DNNFs (SDDExplicit.py)

2018/12/20

    SDD
        - Interaction with the SDD library has been changed to the use of PySDD.
...

2015/05/20 

    LFI learning
        - added option to output model file
    
2015/05/11

    (NEW) Sampling in ProbLog
        - added sampling to the CLI (problog-cli.py sample <model>)
    
    ProbLog CLI
        - added flags for SDD options:
            --sdd-auto-gc / --sdd-no-auto-gc 
                turn on/off minimization and garbage collection on SDD
            --sdd-preset-variables / --sdd-no-preset-variables
                turn on/off initializing the SDD with the correct number of variables

2015/05/06

    ProbLog CLI
        - Logspace evaluation is now the default. Use --nologspace to turn it off.
        
    
2015/05/05

    ProbLog CLI
        - Added a flag to turn on debugging output from grounding engine (--engine-debug)

2015/04/27

    (NEW) ProbLog-to-Python connection
        - added new syntax for calling Python for ProbLog
        
2015/04/26

    ProbLog core
        - eliminate restriction on variable unification

2015/03/05

    ProbLog core
        - add builtin findall/3

2015/02/01

    ProbLog2.1 released. New features include:
    
        - simplified installation
        - better error messaging - improved output
        - added support for weighted model count through Sentential Decision Diagrams
        - several bugfixes
        - possibility to interface with ProbLog's components directly from Python (under development)
        
    Features taken from ProbLog 2.0
        
        - learning from interpretations (learning/lfi.py)
        - weighted model count through d-DNNFs
        - runs on Windows, Mac OSX and Linux
        - ...