1-Machine Learning Fundamentals
This is Lecture 1 of the series. Rather than jumping straight into deep learning, its main goal is to establish the most foundational learning objectives, task types, and training/evaluation workflows in machine learning. For newcomers to this field, this lecture determines whether you can truly understand "where the data comes from, what the model is learning, and how to judge the results" when you later study neural networks.
What This Lecture Covers
This lecture has a fairly complete structure: concepts first, then preparation, and finally a small hands-on case study -- making it ideal for building a broad framework.
- Machine Learning Overview: introduces the definition, historical development, common method taxonomies, typical application scenarios, and real-world challenges of machine learning.
- Machine Learning Preparation: uses the Iris classification task to illustrate what a standard supervised learning problem looks like, and introduces data preprocessing and feature engineering.
- Machine Learning Methods: organizes different learning methods by common tasks such as classification, regression, and clustering, while also covering model evaluation.
- Course Practice: uses Iris classification to tie together the entire pipeline of "data, features, model, and evaluation."
How to Study This
- Focus on problem formulations, not just algorithm names. Start by distinguishing concepts like inputs, labels, features, training sets, test sets, and evaluation metrics.
- When studying the model evaluation section, try to connect terms like accuracy, error, and generalization to specific tasks rather than memorizing them as definitions.
- It is best to do the Iris classification exercise yourself. Even with the most basic tools, this builds more intuition than just reading through the workflow.