实用机器学习

实用机器学习

监督学习

1. 分类
2. 回归

Basic Concepts of Supervised Learning

Sample，example, instance

Features, representations, predictiors

When should we use ML

What is a typical pipline of ML

Define A ML problem -> Construct Dataset -> Transform data &get features -> Design & train a model -> Use the model to predict

A toy example

1. Define a ML problem

   

2. Construct dataset

   

3. Transform data

   

4. Design % train a model(Training)

   

5. Use the model to predict

   

改进

Generalization 泛化

泛化：测试集上的表现

bias:

Variance:

Define a ML problem

无监督学习

1. 聚类
2. 降维