fit op• データ種: simmodel → table
• 呼び出し: import fullseye as fs; fs.ledger.simcal_screen(model=None, method='morris', size=8, threshold=10.0, seed=0) -> 'dict' (実装を直接呼ぶなら import simcal; simcal.simcal_screen(model=None, method='morris', size=8, threshold=10.0, seed=0) -> 'dict'、台帳から引くなら opssimcal.get("simcal_screen"))
どのつまみがカメラの雑音を越えて画像を動かすか、どれが交絡しているか、出力の方向がいくつあるか。
> 以下の詳細説明は原文のままです —— 要約と見出しは訳出済み。
Runs the model on a screening plan over the free knobs' ranges — Morris trajectories (sensitivity.doe_morris, size of them, (k+1) runs each) or a Saltelli plan (sensitivity.doe_saltelli with n = the power of two >= 8 x size rows, (k+2) n runs) — and whitens every output by the noise standard deviation at the initial knobs. The whitened outputs are reduced to principal components and the indices are computed per component with sensitivity.sens_morris / sens_sobol (reused):
Morris mu_star_whitened = sqrt(sum over components of mu*^2) — roughly how many noise standard deviations a full-range move of the knob shifts the images; Sobol' gsi_st = the variance-weighted total index over components (Lamboni et al. 2011). output_dims = number of principal directions of the centred whitened runs above the noise edge (sensitivity.mv_residual_pca with sigma = 1): more free knobs than output dims means the data cannot fix them all. Local confounding at the initial knobs:
sensitivity.mv_confounding on the whitened Jacobian (gamma_K > threshold).
class per knob: no_influence (Morris: mu*_whitened < 1.96, i.e. a full-range move is within noise; Sobol': gsi_st < 0.001), combination (in a confounded group: only a combination is determined), else determined. Needs a noise model (refused without).
• sim_to_real_calibration 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。
• poc_simcal_thin_film — py -3.11 examples/poc_simcal_thin_film.py
• simcal_tour — py -3.11 examples/simcal_tour.py
table を入力に取れる)—
fit)*Provenance: simcal.py — SIMCAL 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.