- Introduced ultrafast communication via sockets to retrieve descriptor scores/fitness from external tools (e.g., python libraries)
- Conversion of molecular representation into DENOPTIM graphs by on-the-fly fragmentation (via GUI or in GA run)
- Extended the functionality of the genetic algorithm by adding optional behavior.
- Generalized SUS selection strategy
- GUI can now handle also settings not directly defined in GUI's forms.
- Minor debugs.
- Removel limitation to max number fo AP and vertex indexes. Now AP and vertex I Ds are unique only within a graph.
- It is now possible to control the maximum size of the subgraphs that the crossover operation can swap. See keyword GA-MaxXoverSubGraphSize.
- Simplified format of symmetric objects inf JSON strings. See src/misc/JSONConverter/convert_symmetric_sets_in_JSON.py for upgradinf existing files.
- Several minor debugs
- Use of Maven to build and test.
- Command line build and test on Windows.
- Creation of single Main method to start any DENOPTIM run or GUI, and open files from command line.
- Introduction of Templates (recursive graphs).
- Refactoring of genetic algorithm.
- Introduction of new mutation operators.
- Introduction of subgraph crossover operation.
- Growth probability can be controlled by molecular size.
- Introduction JSON format for vertexes and graphs.
- Introduction on interface mechanism for interaction with ongoing runs.
- Expanding GUI functionality.
- GA and FSE controlled by GUI.
- Adding internal fitness provider.
- Introduction of GUI
- Keywords changed:
- DenoptimCG:
GA-SortOrder
replaced byGA-SortByIncreasingFitness
, which is a flag that takes null argument.GA-ReplacementStrategy
argument changed to string (NONE or ELITIST).GA-Parallelization
argument changed to string (synchronous or asynchronous).
- DenoptimGA:
GA-FitnessEvalScript
removed. UseFP-Source
andFP-Interpreter
- DenoptimRND:
- FragSpaceExplorer:
FSE-ExternalTask
removed. UseFP-Source
andFP-Interpreter
- DenoptimCG: