About mojimoon

嗨👋欢迎来到 mojimoon 的博客！直达我的 GitHub

• 建立：2023 年 1 月 14 日
• 工具：Hexo
• 服务：Github Pages
• 主题：Stellar
• 加速：jsdelivr CDN

我是一名计算机科学专业的学生，无论前后端开发、数据分析还是机器学习，来者不拒。

参与的一些开源项目：

• 轻小说机翻机器人 - books.fishhawk.top
• 前端：Vite + Vue + Naive UI + Pinia
• 后端：Kotlin + Ktor + Exposed, Docker
• 机翻：GPT (or OpenAI-Compatible API), SakuraLLM
• 主要参与了前端开发，如可展开折叠的章节目录、阅读器样式

• Balatro 模组汇集

• 轻小说文库 EPUB 搜索下载 - mojimoon.github.io/wenku8

• Bangumi 动画排名分析
• 基于 Bangumi15M 构建的偏序网络排名 (Partial Order Network Ranking) - 链接

参与的研究：

1. Ensemble-based Testing for Deep Neural Networks through Scale Unification
   • Proposed a DNN testing and retraining framework based on ensemble metrics that unify different test selection metrics’ scales.
   • Under review at IEEE Transactions on Software Engineering (TSE)
   • Guided by Professor Jacky Keung
2. Empirical Insights of Test Selection Metrics under Multiple Testing Objectives and Distribution Shifts
   • Conducted an empirical study on the effectiveness of various test selection metrics for deep learning systems under multiple testing objectives and out-of-distribution scenarios.
   • Under review at FSE 2026
   • Guided by Professor Jacky Keung
3. Modeling Salary and Job Attractiveness from LinkedIn Descriptions with NLP
   • Implemented a framework combining NLP and machine learning to predict salaries and proposed a Job Attractiveness Index (JAI) based on job description analysis.
   • Guided by Professor Patrick Houlihan at Columbia University