Lecture 1 Linear regression

Linear regression

#This was a lecture of Andrew NG on Couresera website ,  Machine Learning  of Stanford University.#

#This was a note noted by WONG Zinhoo,   Reproduced   please   specify  the Source and the original link #

Question: Price predict.

Size in feet2(x)	Price ($) in 1000s (y)
2104	460
1416	232
1534	315
852	178
…	…

Hypothesis:

The above function is a liner function, with this function; we can present all the line in the 2-demensional plane by changing the parameter

.

The question of linear regression is

How to find the best line’s function to fit the training parameters.

Cost Function

The construction principle of the COST FUNCTION:

The

in the brackets is the hypothesis value of X when the coefficients was

and

. Using this hypothesis minus the target value of Y; we got the distance between the hypothesis line and the training value (target) Y. By changing the value of and (to hypothesis function

are coefficients, while, for Cost function they are arguments); we try to find the line which have minimum distance for all targets. So the function is constructed for:

                     Let

more small,

                    Which equal to:

more small,

                    Which equal to

more small,

                    Which equal to

more small,

If we want to find the minimum value of

we need a derivation of , for computing process more easier we build the   with   .

Linear regression tools

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Linkedin ： http://www.linkedin.com/profile/view?id=307187546

Email :  zihaowang@live.cn

Weibo：  http://weibo.com/mrzihaowang/

Facebook： http://www.facebook.com/zinhoowong

twitter:   https://twitter.com/WongZinhoo