MELV ABM V2.2 — Full Study Dataset: Compound-Gated Cooperative Dynamics in Agent-Based Validation of the Modified Energetic Lotka-Volterra Master Equation
收藏资源简介：
Full 405-run agent-based model (ABM) study for MELV ABM Version 2.2, validating the Modified Energetic Lotka-Volterra (MELV) master equation under compound-gated φ dynamics (Equation 7), the β_norm C1 correction, and i0/i(t) architectural separation. Both the linear form (Equation 1) and saturation form (Equation 1a, η = 0.93) are included. Study design: 81 conditions (3³: ε × φ_init × β_init) × 5 replicates = 405 runs per form; 810 total. 300 generations, 400 agents (20×20 grid), mutation rate σ = 0.005|i0|. Key results: Hartigan dip test: p = 0.00 (< 10⁻³⁰⁰) — strong bimodality confirmed (both forms) ESS invasion recovery: 34/34 = 100% (both forms) Cooperation Theorem: COOP mean CI = 0.9918; COMP mean CI = 0.0273 Pearson r(φ×β, CI_final) = 0.552, p = 1.25 × 10⁻³³ Kruskal-Wallis H (ε groups) = 209.53, p = 3.17 × 10⁻⁴⁶ Compound gating violations: 0 / 810 runs η median recovery (BI-NLS): 0.9300 (planted: 0.9300) Saturation form bounds i(t) > 0 in 100% of runs Relation to V2.1: This dataset upgrades DOI 10.5281/zenodo.19422174 (MELV ABM V2.1) with endogenous φ dynamics, saturation-bounded interaction costs, compound gating, and dual-form testing. The bifurcation threshold I_CRITICAL = 0.9995 is confirmed unchanged. MELVcore preprint: DOI 10.5281/zenodo.19029077MELVcore version: v3.2.0 (Sessions 1–39; 931 tests passing)\nGitHub: https://github.com/NaturesHolismMELV/AIOS
本研究针对MELV ABM 2.2版本开展了完整的405次基于智能体的模型（agent-based model, ABM）实验，旨在验证复合门控φ动力学（公式7）、β范数C1校正以及i0/i(t)结构分离下的改进型能量洛特卡-沃尔泰拉（Modified Energetic Lotka-Volterra, MELV）主方程。本研究同时涵盖线性形式（公式1）与饱和形式（公式1a，η=0.93）两种模型结构。 实验设计：81种实验条件（3³：ε × φ_init × β_init） × 5次重复实验 = 每种形式对应405次运行；总运行次数为810次。实验包含300个世代，400个智能体（20×20网格布局），突变率σ=0.005|i0|。 核心实验结果： 哈特金倾角检验：p值=0.00（<10^-300），证实两种模型形式均存在强双峰性。 进化稳定策略（Evolutionarily Stable Strategy, ESS）入侵恢复率：34/34=100%，两种模型形式均满足该结果。 合作定理：合作组（COOP）的平均置信区间（Confidence Interval, CI）为0.9918；非合作组（COMP）的平均置信区间为0.0273。 皮尔逊相关系数r(φ×β, CI_final)=0.552，p值=1.25×10^-33。 克鲁斯卡尔-沃利斯H检验（ε分组）：H=209.53，p值=3.17×10^-46。 复合门控违规次数：0/810次实验运行。 η值的中位数恢复结果（贝叶斯迭代非线性最小二乘（Bayesian Iterated Nonlinear Least Squares, BI-NLS））：0.9300（预设值：0.9300）。 饱和形式约束下，所有实验运行中i(t)均大于0。 与V2.1版本的关联：本数据集对DOI为10.5281/zenodo.19422174的MELV ABM V2.1版本进行了升级，新增了内源性φ动力学、饱和约束交互成本、复合门控以及双形式测试。分叉阈值I_CRITICAL=0.9995经证实未发生变化。 MELVcore预印本：DOI为10.5281/zenodo.19029077；MELVcore版本：v3.2.0（测试会话1-39，共931项测试全部通过）；GitHub仓库：https://github.com/NaturesHolismMELV/AIOS
