TrAdaBoost: Boosting for Transfer Learning

🤝🤝🤝 Please star ⭐️ this project to support open-source research and development 🌍! Thank you!

This is a teaching and research-oriented project that implements transfer learning using boosting strategies, developed during my stay at Zhejiang Lab (March 1 – August 31, 2023). If you have any questions or need assistance, feel free to reach out!

---

🔬 Overview

Transfer learning aims to leverage knowledge from one or more source domains to improve performance on a target domain with limited data. This project focuses on instance-based methods, particularly variants of the TrAdaBoost algorithm for both classification and regression tasks.

---

📦 Models Included

🔹 Classification

• TrAdaBoost
• MultiSourceTrAdaBoost
• TaskTrAdaBoost
• ExpBoost
• Improved ExpBoost

🔸 Regression

• Transfer Stacking
• TrAdaBoost.R2
• Two-stage TrAdaBoost.R2
• Revised Two-stage TrAdaBoost.R2

Implemented in Python, supporting Windows, Linux, and macOS platforms.

---

📚 Tutorial

• 📘 Tutorial 1: TrAdaBoost
• 📘 Tutorial 2: TrAdaBoost.R2 By Mr. Chen, for AMAT 6000A: Advanced Materials Informatics (Spring 2025, HKUST-GZ). Thanks to Mr. Chen for his valuable contributions!

---

📈 Star History

---

📌 中文介绍（持续更新）

• 📄 TrAdaBoost 推广与介绍

---

📎 Citation

If you use this code in your research, please cite:

Cao Bin, Zhang Tong-yi, Xiong Jie, Zhang Qian, Sun Sheng. Package of Boosting-based transfer learning [2023SR0525555], 2023, Software Copyright. GitHub: github.com/Bin-Cao/TrAdaboost

---

🔧 Package Info

author_email='[email protected]'
maintainer='CaoBin'
maintainer_email='[email protected]'
license='MIT License'
url='https://github.com/Bin-Cao/TrAdaboost'
python_requires='>=3.7'

---

📚 References

1. Dai, W., Yang, Q., et al. (2007). Boosting for Transfer Learning. ICML.
2. Yao, Y., & Doretto, G. (2010). Boosting for Transfer Learning with Multiple Sources. CVPR.
3. Rettinger, A., et al. (2006). Boosting Expert Ensembles for Rapid Concept Recall. AAAI.
4. Pardoe, D., & Stone, P. (2010). Boosting for Regression Transfer. ICML.

---

💡 Related Transfer Learning Methods

1️⃣ Instance-based Transfer Learning

• Instance Selection (same marginal, different conditional distributions): TrAdaBoost

• Instance Re-weighting (same conditional, different marginal distributions): KMM

2️⃣ Feature-based Transfer Learning

• Explicit Distance-based

  • Same marginal, different conditional: TCA (MMD-based) | DAN (MK-MMD-based)
  • Same conditional, different marginal: JDA
  • Both distributions different: DDA

• Implicit Distance-based

  • DANN

3️⃣ Parameter-based Transfer Learning

• Pretraining + Fine-tuning

---

For any inquiries or assistance, feel free to contact Mr. CAO Bin at: 📧 Email: [email protected] Cao Bin is a PhD candidate at the Hong Kong University of Science and Technology (Guangzhou), under the supervision of Professor Zhang Tong-Yi. His research focuses on AI for science, especially intelligent crystal-structure analysis and discovery. Learn more about his work on his homepage.