存量数据复用导向下的FT-MMT方法学导论:低硬件依赖的物质记忆实证路径说明
This paper proposes an exploratory low-threshold research path based on the reuse of stock data, demonstrating the feasibility of conducting research related to the FT-MMT system using public datasets and general experimental equipment without relying on additional large-scale scientific research instruments. Its operational points are guided by directional criteria and do not constitute a finalized standard specification. The study first sorts out the inherent gap of historical evolution information under the traditional instantaneous parameter framework, and clarifies the underlying limitation that hardware accuracy iteration cannot make up for the absence of the historical dimension. Subsequently, it systematically demonstrates the natural adaptability of existing general equipment and global public observation datasets to the retrospective study of material memory effects from three dimensions: microscopic characterization, astronomical observation, and material experiments. This paper sorts out and forms a complete research chain of "data selection - memory coding mapping - superposition fitting comparison - falsifiable prediction output", which only adopts directional criteria throughout the process and does not set absolute quantitative thresholds. This paper does not claim to have obtained complete empirical support, and all inferences are strictly limited to the explorable, verifiable and falsifiable research directions under the FT-MMT research program, providing a unified methodological entry for multidisciplinary cross-comparison and independent verification by external teams. 本文提出一套探索性的低门槛研究路径,依托存量数据复用模式,论证在不依赖新增大型科研设备的前提下,利用公开数据集与通用实验设备开展 FT-MMT 体系相关研究的可行性,其操作要点以方向性判据为准,不构成定型的标准规范。研究首先梳理传统瞬时参数框架下历史演化信息的天然缺口,阐明硬件精度迭代无法弥补历史维度缺失的底层局限。随后从微观表征、天文观测、材料实验三个维度,系统论证现有通用设备与全球公开观测数据集对物质记忆效应回溯研究的天然适配性。本文梳理形成“数据遴选—记忆编码映射—叠加拟合比对—可证伪预测输出”的完整研究链路,全程仅采用方向性判据、不设定绝对量化阈值。本文不主张已获得完整实证支撑,所有推论均限定为 FT-MMT 研究纲领下的可探索、可复核、可证伪的研究方向,为多领域交叉比对与外部团队独立核验提供统一方法学入口。
Authors
- Futao Zhao
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-19
- DOI
- https://doi.org/10.5281/zenodo.24366395
- Primary Topic
- Machine Learning in Materials Science
- Type
- preprint