quantdata.dysanos.grid.gendysanos.conditional#
Time-resolved conditional support diagnostics for generated markets.
- quantdata.dysanos.grid.gendysanos.conditional.run_time_separated_local_tail(*, state, case, log_path_weights, importance_components, split, neighbors, query_paths_per_split, instrument_block_size, seed, device, tail_ranks=(1, 2, 10, 20, 50), signal_thresholds=(2.0, 3.0), progress_interval_seconds=30.0, certification_prerequisites_passed=False, basket_solver_workers=1)[source]#
Fit and value a separate conditional law at every trading time.
Only reference paths enter normalization and nearest-neighbour fitting. Queries are drawn from the direct-P component. A reference query excludes its complete path; because time is never pooled, this is exactly one row.