List view
The `simulated-bifurcation` package will evolve directly from release 1.2.1 to release 2.0.0 as many fundamental and breaking changes occurred over the past months. The purpose of release 2.0.0 will be to provide a more professional and intuitive implementation of the SB algorithm. # Model definition - mathematical definition of models from SymPy polynomials (more natural, more user-friendly, more visual) - optimization domain independent from polynomial: same model can be reused on different domains and no need to separate polynomials in different classes as previously - standardized optimization model API with the `ABCModel` interface # Optimization features - generify optimization domain such that each variable can have its own domain - more efficient stop window that removes converged agents so the computation gets faster - possibility to manually interrupt computation # User interface - make all parameters keyword-only as positional parameters are quite arbitrary - renaming of some parameters for a clearer understanding of their purpose | Previous parameter name | New parameter name | Role | Remarks | | ---------------------------- | ------------------------ | ----- | --------- | | `input_type` | `domain` | Optimization domain(s) of the model | Can also be a list of domains if some variables have its own | | `use_window` | `early_stopping` | Activate agents convergence monitoring | | | `ballistic` | `mode` | Simulated Bifurcation mode (ballistic or discrete) | Would go from boolean value (`True` = ballistic / `False` = discrete) to a string value that can be either `ballistic` or `discrete` | # Documentation - user-friendly documentation website to present the SB algorithm and quadratic optimization more thoroughly - clear docstring
No due date•10/12 issues closed