Glossary
This glossary defines the terms used across public docs, developer docs, package READMEs, and issue work.
Indicators
Indicators are Limen's lowest-level signal helpers. They take a kline-style pl.DataFrame, append one or more derived columns, and return the same frame shape back for the rest of the manifest pipeline.
Features
Features sit one layer above indicators. They combine, re-express, lag, or contextualize price and volume information into model-ready signals or regime flags.
Targets
Target classes in Limen define how the target column is built during the manifest preparation pipeline. A target class is fitted once on the training split, then applied to validation and test without refitting.
Transforms
Transforms in Limen are stateless helpers used during data preparation or target construction. They are not the same thing as train-fitted scalers. For probability calibration and threshold optimization, see limen.calibration.
Scalers
Scalers are train-fitted preprocessing objects. A manifest fits the scaler on x_train, then reuses that fitted state to transform validation and test data without refitting.
Calibration
Calibration controls two model-output surfaces: the reliability of predicted probabilities, and where the decision boundary sits. Limen's calibration layer is declarative — it is configured on the manifest and injected automatically into each experiment round.
Standard Metrics Library
Limen's metrics layer provides the low-level evaluation helpers used inside SFD model functions and reference architectures.
Reference Architecture
The reference architecture is Limen's class-based model layer. It sits underneath the foundational SFDs and underneath Trainer.
Utilities
Utilities are the smaller helper surfaces that support Limen workflows without defining one primary subsystem of their own.