[what is machine learning？]

1.2 [what is machine learning？]

1.人：observation -->  learing  -->  skill

机器：data --> ML --> improved performance measure /skill

2.什么情况下适合使用机器学习：

（1）some 'underlying pattern' to be learned

（2）not easy（programmable） definition ：不是很容易写出一些规则去处理

（3）data about the pattern : inputs

3.example（best suited ML）：

（1）预测婴儿在下一次哪个时间点会哭？  no:  (1)no pattern

（2）判断一个图像中是否包含圆形？  yes   no: (2)很容易写definitioin/program

（3）判断是否给一个用户发放信用卡？  yes  :(1)user behavior (2)not easily program（3）data

（4）地球是否hi在未来十年因为滥用核能而毁灭？ no: (3)no data yet

---

1.3[applications of ML]

1.Food(某家餐厅是否会引起食物中毒)

data:twitter+location

skill:tell food poisoning likeliness of restaurant

2.Clothing

data:sales figures销售数据 + client surveys顾客喜好

skill:give good recommendations to clients

3.Housing

data:characteristics of building and their energy load耗能状况

skill:predict energy load of other buildings closely

4.transportation

data:traffic sign images and meanings交通标志

skill:recognize traffic signs accurately

 5.Education

data:students' records on quizzes on a math tutoring system

skill：predict whether a student can give a correct answer to another quiz question

 

answer correctly~~[recent strength of student > difficulty of question]

data:9 million records from students

ML determines(reverse-engineers)  strength and difficult auto

6.Entertainment

data:how many users have rated some movies

skill:predict how a user would rate an unrated movie

 

data: 1亿 ratings that 480,189 users gave to 17,770 movies(Netflix 线上租赁DVD)

 1.4Formalize the learning problem

input:x->X

output:y->Y

f:X->Y

data: D{(x1,y1),(x2,y2),,,}

hypothesis -> skill  g:x->y

 

{(x n , y n )} from f  -->ML-->  g

A:algorithm

H:hypothesis     利用A从H的众多假设里选择一个最接近f的g.

 

1.5data mining数据挖掘/AI:Artificial Intelligence/Statistics

DM :use huge data to find property that is interesting

ML = DM(KDDCups)

AI：

ML can realize AI,

eg. 下棋：（传统方法：game tree; ML： learning from board data)

Statistics：use data to make inference about unknown process

    g is an inference outcome(预测推论的结果) ；f is something unknown

statistics can be used to achieve ML