（原创）Stanford Machine Learning (by Andrew NG) --- (week 1) Introduction

最近学习了coursera上面Andrew NG的Machine learning课程，课程地址为：https://www.coursera.org/course/ml

在Introduction部分NG较为系统的概括了Machine learning的一些基本概念，也让我接触了一些新的名词，这些名词在后续课程中会频繁出现：

Machine Learning	Supervised Learning	Unsupervised Learning	Regression Problem	Classification Problem	Octave
机器学习	有监督学习	无监督学习	回归问题	分类问题	Octave

---

What is Machine Learning

Definition: A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E.

---

Example of Machine Learning

Suppose your email program watches which emails you do or do not mark as spam, and based on that learns how to better filter spam.

T: Classifying emails as spam or not spam; (目标)

E: Watching you label emails as spam or not spam; (算法 + 数据)

P: The number (or fraction) of emails correctly classified as spam/not spam. (评价方法->损失函数)

---

Supervised Learning

Definition: The goal is, given a labeled training data, to learn a function h so that h(x) is a “good” predictor for the corresponding value of y. A pair (x, y) is called a training example, x denoting “input” variables, also called features, and y denoting “output” or target variable that we are trying to predict.

When the target variable that we are trying to predict is continuous, we call the learning problem a regression problem. When the target can take on only a small number of discrete values, the learning problem is called a classification problem.

A.Example of Regression Problem

Suppose we have a dataset giving the living areas and prices of 11 houses from Portland, Oregon:

Living area (feet2)	Price (1000$s)
450	100
600	140
620	210
...	...

We can plot this data:

So regression problem is to find a function h to fit these points.

B.Example of Classification Problem

Suppose we have a dataset giving the tumor size, patient age and malignant or benign, we plot these data as follows:

So classification problem is to find a function h to sperate these points.

PS: 回归就是找出那个可以拟合样本的函数（平面，空间,...），分类就是找到那个可以把不同类别的样本分开的函数（平面，空间,...）；在特定问题下，比如逻辑回归问题，分类问题就可以被视作回归问题来解决。

---

Unsupervised Learning

In the clustering problem, we are given a training set {x⁽¹⁾, . . . , x^(m)}, and want to group the data into a few cohesive “clusters”. Here, no labels y⁽ⁱ⁾ are given. So, this is an unsupervised learning problem.

PS: 无监督学习很多时候都暗指聚类算法，聚类算法又分硬聚类（K-means, 分层聚类，基于密度的等等）和软聚类（EM算法）。