This repository contains the core implementation of the VCAM module and Oracle Loss as described in our submitted manuscript.
The primary goal of this repository is to facilitate the transition of the View Selection problem from a black-box engineering task to a mathematically grounded research topic.
- Focus on Logic: We have prioritized a clean, modular implementation of our core algorithms. By stripping away redundant engineering wrappers, we ensure that the fundamental logic of VCAM and Oracle Loss is transparent and easy for the community to verify and integrate.
- Core over Code: While the implementation is functional, we believe the theoretical contribution and the view-consistency paradigm are the true keys to this work, rather than the specific software environment used during our multi-phase experiments.
The lead author is currently prioritizing Thesis Writing and Graduation Procedures. As a result, comprehensive step-by-step tutorials are still being compiled.
- What's Available: Functional code for the VCAM architecture and Oracle Loss.
- Future Updates: Detailed environment setups and full training pipelines will be added once the graduation period concludes (estimated May 2026).
- ✅ Core Logic: VCAM module & Oracle Loss implementation uploaded.
- ⏳ Full Pipeline: Documentation, hardware-specific configs, and CLI commands — Coming soon.
本仓库包含投稿论文中 VCAM 模块与 Oracle Loss 的核心实现代码。
本仓库的首要目标是推动 视图选择（View Selection）问题 从复杂的工程任务转向具备数学支撑的研究课题。
- 重逻辑，轻冗余： 我们优先提供了核心算法的模块化实现。通过精简掉繁琐的工程外壳，我们确保了 VCAM 和 Oracle Loss 的底层逻辑清晰透明，便于同行评阅与集成。
- 核心胜于代码： 我们认为，论文提出的理论贡献与视图一致性范式才是研究的关键，而非实验过程中因环境迁移而产生的特定工程脚本。
作者目前正处于毕业论文撰写与答辩准备的最后冲刺阶段。因此，虽然代码“引擎”已在，但“使用手册”仍在整理中。
- 当前提供： VCAM 架构与 Oracle Loss 函数的完整功能代码。
- 后续计划： 待毕业事宜处理完毕（预计 2026年5月），我们将补齐统一的环境封装、预训练权重及完整的端到端运行指南。
- ✅ 核心逻辑： 已上传 VCAM 模块与 Oracle Loss 实现。
- ⏳ 完整流水线： 环境配置 / 运行脚本 / 参数说明 — 后续更新，敬请期待
如有问题欢迎提 Issue，我们会在方便时统一处理与补充说明。