montecarlo op• 資料種類:無 → table(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.mc_photon_slab(layers=None, n_above=1.0, n_below=1.0, n_photons=100000, tol=None, confidence=0.95, batch=50000, seed=1, w_threshold=0.0001, roulette_m=10) -> 'dict'(要直接呼叫實作,import mcphoton; mcphoton.mc_photon_slab(layers=None, n_above=1.0, n_below=1.0, n_photons=100000, tol=None, confidence=0.95, batch=50000, seed=1, w_threshold=0.0001, roulette_m=10) -> 'dict';從台帳取用則 opsconverge.get("mc_photon_slab"))
以 MCML 型蒙地卡羅計算多層混濁平板的漫反射、總穿透與吸收，並附信賴區間。
> 以下的詳細說明為原文 —— 摘要與標題已翻譯。
layers: list of {"n", "mua", "mus", "g", "d"} from the top (coefficients in 1/length, thickness in the same length; d = inf for a semi-infinite last layer). n_above / n_below: ambient refractive indices. A pencil beam enters at normal incidence.
The tracing rules are those of the MCML manual (Wang & Jacques), eqs. 3.2a-3.31 — see the module docstring for which equation does what. Every packet's contribution is kept, so the standard error is the sample standard deviation over packets / sqrt(N) and the interval is mean +- z se (confidence). Batch b uses default_rng((seed, b)).
• tol=None: trace n_photons packets.
• tol=h: sequential Monte Carlo — add batches of batch packets until the half-width of R_d and of T_t is <= h (the stopping rule of converge, from the second batch on), with n_photons as the budget: refusing (NotConvergedError) if the budget is spent first.
Returns Rd, Tt, A (absorbed fraction) and A_layers, each with *_se and *_ci; Rsp (specular, deterministic: eq. 3.2a); energy_balance = R_sp + R_d + T_t + A (1 in expectation — roulette conserves energy only on average, so its deviation is compared with its own standard error); R_total = R_sp + R_d (specular plus diffuse, with the interval of R_d shifted by the deterministic R_sp); n_photons. Defaults: the MCML manual Table 6.1 slab (van de Hulst: R_d = 0.09739, T_t = 0.66096); MCML_TABLE_6_2 holds Table 6.2.
• convergent_computation 族使用指南
• 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。
• converge_tour — py -3.11 examples/converge_tour.py
table 作為輸入)—
montecarlo)*Provenance: mcphoton.py — CONVERGE 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.