Gradient Method
Thm 7.31
給定一個 裡的 symmetric postive matrix ,當我們想解 的 時,等價於找到能 minimizes 的 。
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 →
而怎麼去找 的最小值的方法就叫 Gradient Method
證明:
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 →
而我們利用這個方法去找到每次的 :
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 →
也就是用上一步的 ,加上某一個純量(scalar) 乘上更新的方向
所以我們需要知道 長怎樣,另外一個就是要知道每一步要跨多大,所以要知道 的值是多少,要注意 需要大於 0,這樣才會是我們要的方向
為什叫 Gradient Method
因為之前有說過如果一個函數可微的話,那麼這個函數的負的 gradient 方向就會指出它最大的遞減方向
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 →
所以 就會是 那點的最大遞減方向,那麼 加上 就可以保證越來越小。
這個 是之前說的那個 residual vector,通常定義是
所以這樣我們就可以推出我們下一步的解要沿著 residual vector 的方向來做變化
找 alpha
接下來要來決定 ,推導:
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 →
那我們可以做個簡單的操作來得到 residual vector 的 equation:
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 →
Pseudo Code
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 →
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 →