plan op• 資料種類:無 → design(僅由參數決定的運算子 —— 不接受影像或資料輸入)
• 呼叫: import fullseye as fs; fs.ledger.doe_central_composite(factors=3, alpha='rotatable', center=4, generators=None, inscribed=False) -> 'dict'(要直接呼叫實作,import doe; doe.doe_central_composite(factors=3, alpha='rotatable', center=4, generators=None, inscribed=False) -> 'dict';從台帳取用則 opsmultivar.get("doe_central_composite"))
中心複合設計（Box 與 Wilson 1951）-> design。
> 以下的詳細說明為原文 —— 摘要與標題已翻譯。
Three parts: the two-level factorial (full 2^k, or a fraction from generators as in doe_fractional_factorial — it must be resolution V or better so that every two-factor interaction of the quadratic model stays estimable), 2k axial points at ±alpha on each axis, and center centre points.
alpha: "rotatable" — alpha = F^(1/4) with F the number of factorial points (the prediction variance then depends only on the distance from the centre; Box & Hunter 1957), "face" — alpha = 1 (face-centred, three levels only), "spherical" — alpha = sqrt(k), or a number > 0.
inscribed=True shrinks the whole design by 1/alpha so that the axial points sit at the ends of the range (lo/hi) and the factorial points inside it; otherwise the factorial points sit at lo/hi and the axial points go outside the range — meta['outside_range'] then names the factors whose natural values leave lo..hi (check that the bench can be set there).
• 範例資料目錄(下載 URL / 授權) —— 2-D 用 skimage.data(BSD/公有領域)加合成圖,3-D 給出真實資料源(Stanford/PDS 等)的下載 URL。
• 運算子來歷與參考文獻 —— 該運算子族所依據的研究/方法出處。
• 演算法的正典(作者、年份)與用途見上面的族使用指南。
• doe_plan_tour — py -3.11 examples/doe_plan_tour.py
• multivar_inspection_tour — py -3.11 examples/multivar_inspection_tour.py
design 作為輸入)doe_decode · rsm_fit · doe_main_effects · sens_morris · sens_sobol
plan)doe_lhs · doe_orthogonal_array · doe_fractional_factorial · doe_box_behnken
*Provenance: doe.py — MULTIVAR 運算子登記表。本條目由 tools/opdocs.py md 自動產生(請勿手動編輯)。*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.