Andrew Ng-机器学习-1

1. 什么是机器学习？

google搜索、排序；苹果photo管理功能；邮件过滤等都是机器学习，机器学习是指不根据明确的程序来运行的程序，机器通过学习来获得技能，目前主要是通过神经网络来完成。这是计算机的一种新技能。

Examples:

    -Database mining  从数据中发现新的知识

        Large datasets from growth of automation/web.

        E.g. , Web click data, medical records, biology, engeering

    -Applications can't program by hand

        E.g. , Autonomous helicopter, handwriting recognition, most of Natural Language Processing(NLP)，Computer Vision.

    -Self-customizing programs

    - Understanding human learning

Definition

    -Field of study that gives computers the ability to learn without being explicitly programmed. by Arthur Samuel(1959)

    - Well-posed Learning Problem: 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.

   E--T---->P---T---E better.

Machine learning algorithms:  important-how to use these tools--找到最佳实践

    - Supervised learning.

        回归问题，指预测一个连续值的输出.

        分类问题——离散值的输出。

    - Unsupervised learning

         聚类问题---例如新闻分组，从无序中发现有序 

    - Ohters: Reinforcement leanring, recommender systems.

使用正确的octave，很快的实现算法,好的编程可以很好的节省时间。

2. 几种算法

    - 线性回归  Y=Ax+e 

        Training Set-->learning algorithm--->h(hypothesis-假设)

        如何得到参数值？ 最小二乘法——cost function-->目标函数 goal fuction

        梯度下降算法：outline: start with some A ,keep changing A to reduce goal function.      学习效率(局部最优，不收敛问题--正反馈学习效率 )；梯度更新。

    

转载于:https://my.oschina.net/u/523053/blog/1838009