- 🙃I am a senior student majoring in statistics.
- 🫡This interface stores some of my projects during my undergraduate years.
- 😉Welcome to communicate with me about the content you are also interested in.
|
My main research interest direction is data-driven analysis, starting from specific data tasks to explore effective models and theories. I hope to explore medical auxiliary diagnosis methods based on multimodal data and simpler and more efficient quantifiable analysis strategies through artificial intelligence technology. The main interest directions include multimodal models in the field of machine learning, reinforcement learning, and causal inference in the statistical direction. I really like this short Latin proverb on the right. It is used as the background of my WeChat Moments and has inspired me through many difficult times. I think one day I will be able to find my own rhythm and calm down to do some truly meaningful things！ |
|
As a rough self-introduction, the table of contents of the following content is as follows:
|
|
本项目通过实时屏幕捕捉与OpenCV图像识别技术，结合自适应反应距离算法和鼠标自动化控制，为《汤圆大作战》游戏开发了高效辅助工具。脚本可智能识别加分汤圆、炸弹等道具，动态调整接取策略，在高速移动场景下实现精准操作，最高得分达11000+分，显著超越人工极限。支持ADB投屏控制，包含多级性能优化，助力轻松登顶全行前四~
This project has developed an efficient auxiliary tool for the game "Glutinous Rice Ball Battle" through real-time screen capture and OpenCV image recognition technology, combined with an adaptive reaction distance algorithm and mouse automation control. The script can intelligently recognize props such as bonus glutinous rice balls and bombs, dynamically adjust the pickup strategy, and achieve precise operations in high-speed moving scenarios. The highest score reaches over 11,000 points, significantly exceeding the human limit. It supports ADB screen projection control and includes multi-level performance optimizations to help easily reach the top four in the whole industry.
Link：https://github.com/Augustlnx/Game-Automation-Assistant-with-Computer-Vision
本项目基于贝叶斯方法，针对0-1膨胀负二项回归模型（ZOINB）的参数估计问题展开研究，使用Gibbs抽样与M-H算法对模型参数进行推断，并结合模拟数据验证方法的有效性。
This project is based on Bayesian methods and conducts research on the parameter estimation problem of the zero-one inflated negative binomial regression model (ZOINB). It uses Gibbs sampling and the M-H algorithm to infer model parameters and combines simulated data to verify the effectiveness of the method.
Link：https://github.com/Augustlnx/Statistical-Computing-Course-Project03
|
|
|
本项目利用强化学习Q-learning算法实现一个变体的井字棋游戏。游戏包含经典的 3×3 和 4×3 井字基本策略探索分析，以及考虑引入随机性后如何制定合适的拒绝策略。项目通过对比三种不同的AI算法（Random，MinMax，RL），分析它们在无随机性和有随机性条件下的表现及关于拒绝策略的重要结论，最后提供了完整的实验报告和可交互的游戏界面，供用户体验和研究。
This project uses the reinforcement learning Q-learning algorithm to implement a variant of the tic-tac-toe game. The game includes classic 3×3 and 4×3 tic-tac-toe basic strategy exploration and analysis, as well as considering how to formulate appropriate rejection strategies after introducing randomness. The project analyzes the performance of three different AI algorithms (Random, MinMax, RL) under conditions with and without randomness and important conclusions about rejection strategies. Finally, it provides a complete experimental report and an interactive game interface for users to experience and study.
Link：https://github.com/Augustlnx/Statistical-Computing-Course-Project02
本项目通过理论与实证分析，系统比较了 Bootstrap 和 Jackknife 两种重抽样方法在不同场景下的性能差异，重点关注 小样本条件 和 复杂统计量估计 两类问题。
This project systematically compares the performance differences of two resampling methods, Bootstrap and Jackknife, in different scenarios through theoretical and empirical analysis, with a focus on two types of problems: small sample conditions and complex statistic estimation.
Link：https://github.com/Augustlnx/Statistical-Computing-Course-Project01
本项目是XMU大三上课程《抽样调查》课程设计，主要关注抽样技术和填补方法在调查数据分析中的应用。项目首先涵盖了简单随机抽样、分层抽样、不等概率抽样、整群抽样、系统抽样等多种抽样方法在相同数据集上的估计精度随机模拟结果对比，然后提出了基于Lasso和LightGBM的回归填补方法并进一步比较有效性。
This project is the course design of "Sampling Survey" for juniors at XMU. It mainly focuses on the application of sampling techniques and imputation methods in survey data analysis. The project first covers the comparison of random simulation results of estimation accuracy of various sampling methods such as simple random sampling, stratified sampling, unequal probability sampling, cluster sampling, and systematic sampling on the same data set. Then, it proposes a regression imputation method based on Lasso and LightGBM and further compares its effectiveness.
Link：https://github.com/Augustlnx/Sampling-Survey-Course-Project
本项目是XMU短学期课程《博弈论》课程论文——关于博弈论在沙滩占位模型中的应用研究。项目探讨了在不同顾客分布假设下，多个小贩如何通过选择位置来最大化顾客数量，并分析了不同场景下的纳什均衡策略。
This project is the course paper of "Game Theory" in the short semester course of XMU - Research on the application of game theory in the "beach occupancy model". The project discusses how multiple vendors can maximize the number of customers by choosing locations under different customer distribution assumptions, and analyzes the Nash equilibrium strategy in different scenarios.
Link：https://github.com/Augustlnx/Game-theory-Course-Project
这是一个XMU大二下课程《应用随机过程》 的课程论文，本文只回顾并详解了经典书籍 Essentials of Stochastic Processes (Rick Durrett .2ndEdition) 和 STOCHASTIC PROCESSES (Ross .2ndEdition) 中提到的大富翁游戏的经典例题，并通过马尔科夫链和离出概率分析了一些游戏中的简单情形下的问题，并最后通过鞅理论探讨了一种简化的资金流动模型和破产时间估计。
This is a course paper for the course "Applied Stochastic Processes" in the second semester of sophomore year at XMU. This paper only reviews and elaborates in detail the classic example of the "Monopoly game" mentioned in the classic books "Essentials of Stochastic Processes (Rick Durrett. 2nd Edition)" and "STOCHASTIC PROCESSES (Ross. 2nd Edition)". And it analyzes some problems in simple situations in the game through "Markov chain" and "escape probability", and finally discusses a simplified capital flow model and bankruptcy time estimation through "martingale" theory.
Link：https://github.com/Augustlnx/Stochastic-Processes-Course-Project