An Introduction to Machine Learning: Concepts, Learning Path, and Knowledge System

Author Wang Maolin from Huazhong University of Science and Technology, sourced from Datawhale, shares a structured guide to help beginners enter the field of machine learning.

Part 1: Machine Learning Related Concepts

Artificial intelligence (AI) encompasses many terms such as machine learning, statistical learning, data science, data analysis, data mining, and deep learning. The article first clarifies the difference between statistics and statistical learning, noting that statistical learning draws heavily on statistical foundations.

Machine learning is defined as a set of fixed algorithms or methods that enable computers to learn rules or mappings from data; it originated from statistical learning, making the two concepts largely equivalent except for deep learning.

Data science and data analysis are briefly described, and the relationship between statistical learning, machine learning, and deep learning is illustrated.

Part 2: Overall Machine Learning Learning Process

The learning process is divided into three interrelated blocks: mathematical foundations, machine learning algorithms, and programming practice. The article emphasizes that these blocks are not strictly sequential; learners can start with algorithms or programming and supplement mathematics as needed.

Mathematical Foundations

Key topics include calculus derivatives, linear algebra matrix operations, probability theory (total probability, conditional probability), information theory (entropy), and optimization methods such as gradient descent and KKT conditions. Advanced topics for research include learnability, complexity, generalization, and variational methods.

Machine Learning Algorithms

Algorithms are split into traditional machine learning (e.g., linear regression, logistic regression, decision trees) and deep learning (ANN, CNN, RNN). Each sub‑model can be studied independently.

Programming Practice

Python is recommended as the most accessible language for machine learning. Learners should start with basic syntax, then explore NumPy, Pandas, followed by the scikit‑learn framework for traditional algorithms and Keras for deep learning, practicing alongside theory.

Part 3: Machine Learning Knowledge System

The article proposes a three‑part knowledge system—machine learning theory, algorithms, and practice—and lists recommended resources and Datawhale projects for each area.

Key Book (theory supplement): https://github.com/datawhalechina/key-book

Pumpkin Book (detailed study of Zhou Zhihua's book): https://github.com/datawhalechina/pumpkin-book

Easy‑RL (deep reinforcement learning tutorial): https://github.com/datawhalechina/easy-rl

LeeML‑Notes (Li Hongyi’s machine learning course notes): https://github.com/datawhalechina/leeml-notes

The article concludes by thanking readers and encouraging them to like, share, and join the DataFunTalk machine learning community.