Zhou Zhihua Machine Learning (Watermelon Book)
Descriptions
- Source: Zhou Zhihua (Nanjing University)
- Prerequisites: Linear algebra, probability, calculus
- Difficulty: 🌟🌟🌟
- Class Hour: 60+
Zhou Zhihua's Machine Learning (nicknamed the "Watermelon Book") is the most classic ML textbook in China. The name comes from the author's use of watermelon data as a running example throughout the book.
The book covers core ML algorithms and theory, comprehensive but theory-heavy. It's suitable for systematic ML theory study but not for complete beginners.
Resources
- Book: Douban
- Companion: Pumpkin Book (detailed formula derivations)
- Lectures: Bilibili
Content Structure
Part 1: Foundations
- Model evaluation and selection
- Linear models
- Decision trees
Part 2: Core Algorithms
- Neural networks
- Support Vector Machines
- Bayesian classifiers
- Ensemble learning
Part 3: Advanced Topics
- Clustering
- Dimensionality reduction
- Feature selection and sparsity
- Computational learning theory
Self-Study Tips
Use with the Pumpkin Book. The Watermelon Book's formula derivations are brief; the Pumpkin Book provides detailed derivations.
Read after Andrew Ng. The Watermelon Book is theory-heavy. Build intuition with Andrew Ng's course first.
Do the exercises. Each chapter has exercises to test your understanding.
Notes
The Watermelon Book is a reference for ML graduate entrance exams and interviews in China. Essential for systematic ML theory study.
However, it's not suitable for complete beginners. Start with Andrew Ng's course, then read the Watermelon Book.