API Reference

The API reflects the two-stage pipeline described in the paper: an Extractor turns a dataset into a meta-feature vector, an Evaluator scores circuits to build ground-truth labels, and a Recommender predicts a circuit ranking from meta-features. The Inference entry points connect them together; the Circuits module holds the encoding-circuit pool and kernel computation.

Page	Contents
Inference	recommend, preprocess_new_dataset, evaluate_recommendation - the end-to-end entry points
Extractors	get_extractor, BaseExtractor, ExtractionResult, task-specific extractors
Evaluators	get_evaluator, BaseEvaluator, task-specific Oracles, filter_degenerate_datasets
Recommender	load_default_recommender, get_recommender, PairwiseRecommender, model-selection utilities
Circuits	CIRCUIT_POOL, UNIT_RANGE_CIRCUITS, kernel-matrix computation