Lecturer · MSc in Computer Science (Data Science), Universiti Malaya
静态横幅 / Static banner · 原创动画 / Behind the animation
Engineering & research portfolio · Engineering highlights · Research brief · Academic CV · Learning notebook · guhang2011@126.com
I'm Hang Gu (顾航), a lecturer and teaching section director at 江苏财经职业技术学院, with an MSc in Computer Science (Data Science) from Universiti Malaya and a background in frontend engineering.
Engineering focus: frontend systems, explicit workflow states, data-quality checks and verifiable delivery. The public cases connect implementation choices with code, interactive demonstrations and scoped verification records.
Research direction: multimodal learning and intelligent systems engineering. My proposed doctoral work studies reliable multimodal emotion recognition with limited labels, degraded inputs and expanding classes.
这里以工程技术为主：前端交互、工作流状态、数据质量与可复查交付。通过公开代码、项目演示和学习笔记说明设计取舍；同时保留有限标注下的多模态学习与模型评估研究记录。
27 October 2026 update / 研究更新： 第一作者 / First author — Hybrid AI Framework for Multi-Hazard Climate Risk Governance in Renewable Energy-Water Nexus Systems，已投稿 Climate Risk Management，尚未录用 / submitted, not accepted。围绕中国 30 个省份（2005—2024）的能源—水系统，以 TFT–GNN 结合时间建模与跨区域风险传播。双语研究介绍与方法示意 / Study overview。
24 October 2026 update: the revised manuscript has been submitted to Neurocomputing and is not accepted. New title: Where do fusion gains come from in low-resource multimodal fake news detection? Operating points, archive artifacts and frozen CLIP baselines. Study overview · Updated doctoral research brief.
研究主线： 从融合收益审计，走向资源受限交互中的可靠情绪识别。2026 年 10 月 25 日依据 RP 更新：三项方法研究递进展开，系统验证贯穿全过程；以下均为拟研究内容。
| 拟研究 / Proposed study | 要回答的问题 / Question | 验证依据 / Planned evidence |
|---|---|---|
| 01 预算约束适配 · Budget-aware adaptation | 复杂融合是否值得额外成本？ | 训练、选模、校准统一预算；Macro-F1、UAR、配对重复与成本 |
| 02 证据感知决策 · Evidence-aware decisions | 证据不足时何时请求复核？ | 质量、分歧、校准消融；相同风险或覆盖率下的比较 |
| 03 遗忘约束更新 · Continual adaptation | 学习新类后旧类与置信质量能否保持？ | 固定回放；新旧类表现、遗忘、校准漂移与更新开销 |
| 04 系统综合验证 · Integrated validation | 受损输入与运行异常下整套方法是否一致？ | 独立测试、端到端延迟、复核触发与版本回退 |
Read the bilingual research plan / 阅读双语研究计划 · Protocol checklist / 实验协议清单
工程基础如何支撑研究： 电信开户系统中的动态表单、状态机与接口契约，弱网应用中的离线缓存与恢复同步，为拟研究原型的缺失输入处理、异常恢复和版本一致性提供工程基础。公开状态机示例用于说明相关能力，属于独立学习演示。
| Research question | Approach | Authorship | Recorded status |
|---|---|---|---|
| Climate risk in renewable energy–water systems | TFT + GNN-CRP; temporal dependence and interregional propagation; 30 Chinese provinces, 2005–2024 | First author / 第一作者 | Submitted to Climate Risk Management; not accepted / 尚未录用 |
| Where do fusion gains come from? · Low-resource multimodal fake-news detection | Reanalysis of 344 archived training runs: frozen-CLIP baselines, decision operating points and archive artifacts; FCP is studied, not claimed as a new architecture | First author | Revised and submitted to Neurocomputing; not accepted |
| Validation sensitivity in limited-label news classification | Model selection and paired validation sampling | Order not stated in CV | Submitted to DASFAA 2027 |
| Vision-language multimodal behavior prediction | VideoMAE, DistilBERT, cross-attention and LSTM temporal modeling | First author | Accepted at ISCAIT 2026 |
The Neurocomputing entry follows the revised manuscript and my status update on 24 October 2026. The Climate Risk Management entry follows my abstract and first-author confirmation of 27 October 2026. Other entries follow the supplied CV. The homepage update does not rerun research experiments.
My MSc thesis, Dental Deformities Prediction Using Bayesian Networks among Adolescents, used R and Bayesian networks to study conditional relationships in clinical data.
Research brief · Discussion draft · Code & verification evidence · Full academic CV
移动优先的老乡会员社区，支持会员与版主管理、活动定位签到、资源互换、图片／文件聊天、论坛语音和家乡特产订单；天气支持设备定位与网络城市估算。基于 Vue 3 · uni-app · Spring Boot，提供电脑网页、Android / iPhone H5 与微信小程序源码。
申博作品集 / Portfolio · 源码与部署 / Source & setup · 完整图集 / Screenshots · 语音功能 / Forum audio
验证 / Verification: 60 项后端测试、17 个共享异步场景通过；Web、H5、小程序构建通过。当前为本地试点版本，小程序待真机验收与发布。语音截图使用合成测试数据。验收记录 / Validation scope。
天气选城与网络定位截图 / Weather & location · 会员欢迎界面 / Membership entry
早期工程项目整理 / Earlier frontend work: 生产反馈 PC 端、移动端、思悦后台与电信固网平台 — 从旧简历整理业务场景、本人职责和技术栈；关联 Demo 为独立学习示例。GitHub 项目文档。
| Work | Focus and scope | Read / verify / try |
|---|---|---|
| Shuyang Home · 沭阳之家 · Community engineering | Mobile-first Vue / uni-app community; membership, location check-in, moderated voice posts, private attachments and offline-settled specialty orders | Source & setup · Screenshots · Validation |
| KICP · Research | CLIP, learned prompts, co-attention and knowledge retrieval for multimodal fake-news detection | Code & run requirements |
| KAPT · Reproducibility learning | Curated KANO-based prompt components, validated early stopping and sample-aligned probability ensembles; synthetic examples | Reading & source attribution · Reproduction · Early-stopping replay |
| Silk Road Intelligence · Teaching engineering | Cross-border order workflow, stock consistency, permission scopes and simulated fulfillment; adapted from an existing college application | Screenshot gallery · Engineering notes |
| Counselor Workflow · Engineering case | Role-based interaction, state transitions, version checks and retry contracts; curated static demonstration | Architecture & scope · Online demo |
| Frontend & Data Lab · Learning experiments | Synthetic-data analytics, annotation agreement, bounded crawling and model-evaluation notes | Reading map · Interactive lab |
| React Workflow Board · React learning project | Typed state transitions, version conflicts, retry paths and keyboard-readable controls; synthetic tasks | Live React app · React source · Project notes |
| Schema-driven Form Studio · Low-code engineering example | Schema / renderer separation, field ordering, explicit validation and JSON inspection; synthetic request form | Online studio · Schema · Design notes |
| Reliability Gate Lab · Multimodal reliability learning project | Synthetic quality, confidence, disagreement and missing-modality controls; shows auto-output, review and insufficient-evidence states | Online lab · Learning note · Research question |
Reading path: research question → core code → verification steps and limitations → interactive case. Model self-study notes.
笔记集现有 19 篇，新增以下四篇：
可运行的 Python 示例与 10 项测试。合成教学数据；不作为论文的实验结果。
| Capability | What is visible in the public work |
|---|---|
| Interface contracts | Dynamic forms, role-scoped states, keyboard access, responsive layouts and browser-level checks. |
| Data quality | Deterministic cleaning, annotation agreement, quarantine records, CSV/SQLite/JSON hand-offs and bounded crawling. |
| Model evaluation | Baselines, validation sensitivity, threshold selection, confusion analysis and sample-aligned prediction ensembles. |
| Evidence-shaped delivery | Release gates, artifact hashes, scope notes, synthetic fixtures and explicit “not rerun” boundaries. |
工程阅读顺序： 需求问题 → 输入与状态契约 → 核心实现 → 定向检查 → 范围与限制。主页的 Engineering lens 将这些能力映射到项目案例、在线实验和验证记录。
18-second project walkthroughs / 项目动画： Early stopping · 模型早停 · Data quality audit · 数据质量审计. Silent animations generated from the demo logic with synthetic data; captions and interactive experiments are available on the site.
The learning lab is a newly assembled educational example. It is separate from employer software and research experiments. KICP contains research code; its README describes the external data and assets required to run it. KAPT preserves KANO source attribution and documents the scope of its component tests.
Earlier coursework / 早期课程练习
- WQD7006: historical Python exercises in data collection, regional aggregation and visualization.
- introdatascience: early R exercises in extraction, data cleaning and JSON serialization.
丝路智链 · 桌面与手机界面 / Silk Road Intelligence gallery
从既有学院应用改造的跨境订单教学案例，展示中哈合作主题登录、营销审核、订单台账与物流仿真。下列业务界面使用合成教学数据。
 |
 |
既有后端报告记录 41 项通过（含共享账户基础测试），本次未重跑。支付与设备为仿真，来源和验证范围详见案例。
把代码中的问题整理成可以复查的笔记：为什么这样设计、如何验证、还有什么没有回答。在线阅读 19 篇双语笔记，支持主题筛选、搜索和打印。
| 主题 / Topic | 阅读入口 / Read |
|---|---|
| 近期学习 / Latest | 证据感知多模态决策 · 总标注预算 · 模型版本回退 |
| 模型学习 | 早停、样本对齐与集成 · 提示学习的张量契约 |
| 前端与数据 | 双语搜索、焦点与动态效果 · 数据清洗的质量契约 |
| 项目心得 | 订单幂等与库存一致性 · 权限边界与可复查交付 |
| 生成式模型与工程复现 | 数字人技术路线 · 文生图原理与操作 · 版本、提示词与证据链 |
Collection updated on 27 October 2026 from public code and project documents, with AI-assisted drafting and translation. Each note separates source observations, teaching examples and proposed checks, with links to evidence; these are not new benchmark results or a historical lab diary.
从本地高校资料归档与协作项目整理的原创笔记，记录设计取舍、验证结果和待完成的工作。附上线、容量测试与个人 AI 接入模板。
全部 6 篇笔记 / All notes · 官方资料导航 / Official reading
Design decisions, validation and reusable checklists / 设计取舍与验证
| Start with a question | Read / Use |
|---|---|
| “500 人在线”到底验证了什么？ | 容量复盘 · 测试报告模板 |
| 每个账号如何使用自己的 AI？ | 密钥与模型接入 · 验收表 |
| 从开发机到学校服务器还差什么？ | 部署笔记 · 发布检查表 |
These are original notes and reusable checklists, with explicit testing limits. Linked references stay on their official sites; private product code and user data are not published here.
上传系统界面预览 / Upload system preview
来自本地 Edge/WebKit 自动化测试夹具的合成界面截图，仅用于展示布局和功能状态，不包含真实账号、API Key 或用户文件。
 |
 |
 |
Applied experience and teaching / 工程与教学经历
- Intelligent applications: document parsing, entity extraction, semantic retrieval and RAG workflows.
- Frontend systems: Vue / React, TypeScript, dynamic forms, interface contracts, accessibility and performance.
- Data systems: Python, R, SQL, Hadoop / Spark, reproducible analysis and visualization.
- Teaching: computing courses and practical projects spanning data collection, storage, analysis and presentation.
Recent applied work includes an ongoing AI contract-management project, an ongoing intelligent information-retrieval project, and a completed modular administrative-dashboard project. These are described as experience; their proprietary source code is not part of this profile.
Profile updated 28 October 2026; Shuyang Home has been added with mobile screenshots, forum audio and scoped validation records; the homepage now also includes a synthetic reliability-gate lab and three notes extending the proposed research protocol. The Neurocomputing entry follows the revised manuscript and the author’s latest status update; other research states follow the supplied CV; see the portfolio for evidence and scope.