Exploration 🚀

A collection of small exploration projects in AI, Data Science, and Documentation.

---

About

This repository contains various experimental projects exploring different aspects of artificial intelligence, machine learning, data science, and technical documentation. Each project is a self-contained learning experience focusing on practical implementation and experimentation.

---

Projects

🔤 TransformerPractice

Machine Translation with Transformer: Full Process and Implementation

A complete end-to-end implementation of the Transformer architecture for neural machine translation, built from scratch.

• Tech Stack: PyTorch, torchtext, Multi30K Dataset

• Key Features:

  • Complete Transformer architecture implementation
  • Multi-head attention mechanism
  • Positional encoding
  • BLEU score evaluation system
  • Learning rate scheduling
  • Jupyter notebook for pipeline analysis

• Quick Start:

  cd TransformerPractice
  conda env create -f environment.yml
  python main.py --epochs 1000 > output/output.log

• Documentation: View Details

---

📊 GraphPlotting

Data Processing and Visualization Tool

An accumulated graph plotting tool that integrates data preprocessing and visualization capabilities for quick experiment result presentation.

• Tech Stack: Pandas, NumPy, Matplotlib, Seaborn

• Key Features:

  • Data loading (Excel, CSV support)
  • Data cleaning and preprocessing (interpolation, normalization)
  • Multiple standardization methods (Z-score, Min-max)
  • Histogram and KDE comparison plots
  • Custom fonts and styling options
  • Batch graph generation and archiving

• Core Components:

  • DataProcessor: Handle data loading, filling missing values, and standardization
  • GraphPlotTool: Create histogram plots, KDE plots, and comparison visualizations

• Use Cases: Robust Prompt Learning analysis, training progress visualization

• Documentation: View Details

---

📝 DocsWriting

Documentation and Writing Templates

A carefully crafted collection of document templates for academic presentations, daily logging, and documentation writing.

• Tech Stack: Typst, LaTeX, TikZ, tcolorbox

• Projects Included:

  1. Concise Slides (Typst)

  • Minimalist slide template based on Touying Dewdrop theme
  • Clean design with background image support
  • Three-column footer layout
  • Predefined color schemes
  • For: thesis defense, research reports, course presentations

  2. One-sentence Log (LaTeX)

  • Daily logging template with visual hierarchy
  • Automatic month-based color system
  • Task completion tracking with auto-numbering
  • Hierarchical goal management
  • Beautiful concept boxes and dividers
  • For: work logging, learning tracking, goal management

• Quick Start:

  # For Concise Slides
  cd DocsWriting/ConciseSlides
  typst compile slide.typ
  
  # For One-sentence Log
  cd DocsWriting/One-sentence-Log
  xelatex poster.tex

• Documentation: View Details

---

Tech Stack

Deep Learning: PyTorch
Data Processing: Pandas, NumPy
Visualization: Matplotlib, Seaborn
Typesetting: Typst, LaTeX, TikZ

---

References

• TransformerPractice:

  • hyunwoongko/transformer
  • Transformer: Concept and code from scratch

• DocsWriting:

  • Touying Framework
  • Typst

---

Author

Zhihao Li
Homepage: https://lzhms.github.io/

---

License

This project is licensed under the MIT License.