sensitivity op• データ種: なし → design(引数だけで決まる op —— 画像やデータの入力を取らない)
• 呼び出し: import fullseye as fs; fs.ledger.doe_morris(r=10, factors=4, levels=4, seed=0, n_candidates=0) -> 'dict' (実装を直接呼ぶなら import sensitivity; sensitivity.doe_morris(r=10, factors=4, levels=4, seed=0, n_candidates=0) -> 'dict'、台帳から引くなら opsmultivar.get("doe_morris"))
Morris one-at-a-time screening plan: r trajectories of k+1 runs -> design.
Each trajectory starts at a random point of the levels-level grid on the unit cube and moves one factor at a time by delta = levels / (2 (levels - 1)) (Morris 1991; with an even levels every grid level is equally likely to be visited), so r (k+1) runs give r elementary effects per factor. n_candidates > r draws that many trajectories and keeps r of them that are spread out: the trajectory distance of Campolongo et al. 2007 (sum over point pairs of their Euclidean distances), chosen greedily — a pair of the farthest trajectories first, then repeatedly the trajectory that adds the most distance (an approximation of their exhaustive search, stated in meta['selection']).
The coded matrix is 2u - 1 of the unit-cube points; meta keeps the trajectory bookkeeping that sens_morris needs (changed factor and sign per step).
• design_of_experiments ファミリ ガイド
• サンプルデータ カタログ(DL URL / ライセンス) — 2-D は skimage.data(BSD/public)+ 合成、3-D は実データ源(Stanford/PDS 等)の DL URL。
• 演算子の来歴・参考文献 — この op 族の元になった研究/手法の出典。
• アルゴリズムの正典(著者・年)と用途は上記ファミリ使い方ガイドに記載。
• sensitivity_tour — py -3.11 examples/sensitivity_tour.py
design を入力に取れる)doe_decode · rsm_fit · doe_main_effects · sens_morris · sens_sobol
sensitivity)sens_morris · doe_saltelli · sens_sobol
*Provenance: sensitivity.py — MULTIVAR operator registry. この per-op ノートは tools/opdocs.py md が自動生成(手編集しない)。*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.