33  Object structure

The public workflow uses functions and typed S7 objects. Configurations hold inputs; fitted results hold the outputs of an executed operation.

flowchart LR
  H[Hyperparameters] --> T[train]
  P[SupervisedPreprocessorConfig] --> T
  D[DecompositionConfig] --> T
  U[TunerConfig] --> T
  R[ResamplerConfig] --> T
  E[ExecutionConfig] --> T
  T --> S[Supervised or SupervisedRes]
  S --> V[review and writeup]
  S --> J[record and write_result]

Regression and Classification specialize Supervised. RegressionRes and ClassificationRes specialize SupervisedRes. HardClustering and SoftClustering distinguish assignments from assignments with a membership matrix. Decomposition retains the fitted transformation and component scores.

Use @ to access a declared S7 property, such as model@hyperparameters. Use [[ for ordinary lists. Setup functions validate configuration objects; serialization uses their published contracts. See records for the distinction between a recipe, a run record, and a portable result.