Watermelon book / Pumpkin book -- Chapter 1 and 2 Summary

Catalog

  Chapter one   The introduction

1. Basic terminology

 2. Hypothetical space

3. Generalize preferences

Chapter two   Model evaluation and selection

1. Empirical error and over fitting

 2. Evaluation methods

 3. Performance metrics

---

  Chapter one   The introduction

This chapter mainly introduces some basic concepts of machine learning , Including basic terms , Hypothetical space , Inductive preference, etc ;

1. Basic terminology

Data sets ：

In terms of characteristics ： It is divided into input space and output space ;

From the sample ： Divided into training sets , Test set ; The samples of training set can be divided into training set and verification set ;

From the type of machine learning ： It is divided into supervised learning and unsupervised learning ; Supervised learning is labeled , Have output , Each piece of data is also called an example .

 2. Hypothetical space

Suppose that space is simply a set of input spaces .

3. Generalize preferences

What I understand is the fitting effect of the model .

Occam's razor principle ： If there are more than one hypothesis consistent with observation , Just choose the simplest one ; let me put it another way , If multiple models can achieve better results , Just choose the simplest model ;

There is no free lunch Theorem （NFL Theorem ）： When all opportunities are the same or all problems are equally important , The expected performance of all learning is the same .

Chapter two   Model evaluation and selection

Model evaluation is very important for judging the performance of a model , This part mainly includes empirical error and over fitting 、 Evaluation methods 、 Performance metrics .

1. Empirical error and over fitting

Over fitting ： The model is too complex ;

Under fitting ： Model learning ability is low .

 2. Evaluation methods

 

 3. Performance metrics