Custom graph model optimization
To achieve to applicability of the gp_optimizer to structural learning of bayesian networks, the following changes are implemented:
- added GraphObject - base class for Chain, as well as GraphNode/PrimaryGraphNode/SecondaryGraphNode
- structural DAG-related methods moved to GraphNode
- simplified example of custom model learning
- constraints customization implemented
Now gp_optimizer can be used for any graph-based optimization problems.
Probles:
- GraphObject cannot be added to the history