Machine learning notes

summary ：

machine ： By means of acquired learning , Acting very wise , Today's Machine Learning：you write the program for learning.

Machine Learning:Looking for a Function From Data.

The ultimate goal ： When the machine has the ability to learn , By teaching children , Teaching machine

Steps of machine learning ：

1. First of all, there are thousands of function, That is, the set of functions , This set of functions is called Model （Model）

2. Let the machine measure function The stand or fall of （training data）

3. There is a good algorithm , From many function Pick out the best one , For later use , The purpose of this ： Make the machine run efficiently
 

Ask questions before you learn ：

1. In the mind map below , What needs to be solved task（ Mission ）, What are the problems to be solved model, There are four main ways of machine learning The characteristics and specialties are what ？

2.Supervised Learning and Reinforcement Learning What are the applicable scenarios ？

Learning Map

Professor Li Hongyi Learning Map Exhibition ：

Text introduction ：

1.regression—— regression analysis

The output of the target function f is "scalar".

Find a function , Output the prediction result of a numerical class ;

example ：PM2.5 The forecast ,

The food is from the past PM2.5 Training materials for , The output is today's PM2.5 The numerical .

2.classification—— classification

        2.1 Binary Classification    output：Yes or No

        2.2 Multi-class Classification     output:  class1 class2 class3 ...

explain ：Deep Learning Of function Extremely complex , So you can do something very complicated , Such as speech recognition , Image recognition ,Playing go（ go ）

3. The way machines learn skills ：

Supervised learning That is, learning under the guidance of human beings , At the same time, the machine should have the ability to draw inferences from one instance , It's the same as the teacher telling you how to do ;

Reinforcement Learning Autonomous machine learning , Reflect on learning from practical evaluation , Machines are required to have high intelligence , At the beginning, I didn't have a lot of original materials , It's the evaluation system ; stay Reinforcement Learning In the process of , Due to efficiency requirements , The opponent of a machine is often another machine , Such as AIphaGo, Is to make continuous progress in the process of playing chess with yourself ;

4. Doing it regression The difficulties you will encounter

Statement ： This note is used by the author in the process of learning online class with teacher lihongyi , Please use dialectically , Ongoing update ...