Engine reference

The layers behind the public API, for contributors. _core holds the pure values: the option aliases, the defaults, the validators, and the containers (penalties, stages, the scaling map, the summary records). _internals holds the mechanism: the torch module with its heads and recursions, the objective, the staged trainer, the predictor that runs the trained network on numpy data, the fit record, the model base, and the behaviour mixins the result composes. The public modules appear a second time below with each class’s own private members. Everything on these pages may change between releases without notice.

Private layers

ns_sdn.engine._core	Primitive layer: constants, value objects, validators, and type aliases.
ns_sdn.engine._internals	Mechanism layer: the network, the trainer, the predictor, the family base.

Public surface, private members

ns_sdn.spectral_decomposition	The Non-Stationary Spectral Decomposition Network: specification and result.
ns_sdn.spectral_state	Spectral-state coordinates: the decomposition read as a time-local measure.