最小平方逼近多項式(Polynomials of Least square)
目的
我們想要用一個多項式來逼近另一個 function ,這個多項式我們寫成
這樣的話 least square error,或一開始的 LDA 的 error 就會長 ,那一樣,我們要找到 來最小化 :
Image Not Showing
Possible Reasons
- The image file may be corrupted
- The server hosting the image is unavailable
- The image path is incorrect
- The image format is not supported
Learn More →
推導
而我們要最小化這個 就要用到 gradient 了,也就是說 ,或妳也可以說
那我們就可以開始推了:
Image Not Showing
Possible Reasons
- The image file may be corrupted
- The server hosting the image is unavailable
- The image path is incorrect
- The image format is not supported
Learn More →
因為 A 是個 ill-condition 且稠密的矩陣,如果要解這個線性系統會很麻煩,非常沒有效率,因此我們就要換個建構多項式的方法,其中一種方法就是利用線性獨立來操作
在操作之前要先複習一個概念:一個多項式的集合 線性獨立 iff
那我們假設 是一個 degree 為 j 的多項式,那麼 在任何區間 上都會線性獨立,因為他們 degree 不同,像是 和 就線性獨立
所以現在 ,那一樣我們要找 、 等係數來最小化 :
Image Not Showing
Possible Reasons
- The image file may be corrupted
- The server hosting the image is unavailable
- The image path is incorrect
- The image format is not supported
Learn More →
然後一樣找 gradient E = 0:
Image Not Showing
Possible Reasons
- The image file may be corrupted
- The server hosting the image is unavailable
- The image path is incorrect
- The image format is not supported
Learn More →
例子
Example 1. 勒壤得多項式 Legendre Function
Image Not Showing
Possible Reasons
- The image file may be corrupted
- The server hosting the image is unavailable
- The image path is incorrect
- The image format is not supported
Learn More →
那個 、… 是我們取的
Example 2. 柴比雪夫多項式 Chebyshev polynomials
Image Not Showing
Possible Reasons
- The image file may be corrupted
- The server hosting the image is unavailable
- The image path is incorrect
- The image format is not supported
Learn More →
那個 、… 是我們取的