--- tags: Linear Algebra - HungYiLee --- Linear Algebra - Lec 36: Singular Value Decomposition === [TOC] # [課程網站](http://speech.ee.ntu.edu.tw/~tlkagk/courses_LA18.html) # [Lecture 36: Singular Value Decomposition](https://www.youtube.com/watch?v=OEJ0wxxLO7M&list=PLJV_el3uVTsNmr39gwbyV-0KjULUsN7fW&index=37) ## Outline  ## SVD   - $\Sigma$ 長的就是右上角那樣,且 $\sigma_1\geq\sigma_2\geq...\geq\sigma_k\geq 0$ **$A$ 的 rank 就是 $\Sigma$ 的 rank**,因為 - 我們可以知道 $Rank(\Sigma)=k$ - review Lec 12?,可以知道 $Rank(AB)\leq\min(Rank(A),Rank(B))$ Rank 部分 review Lec 12? - 又 $Rank(U)=m, Rank(V^T)=n$ - 且我們知道 - If $A$ is a $m\times n$ matrix, and $B$ is a $n\times k$ matrix, then $Rank(AB)\leq\min(Rank(A),Rank(B))$ - If $B$ is a matrix of rank $n$, then $Rank(AB)=Rank(A)$ - If $A$ is a matrix of rank $n$, then $Rank(AB)=Rank(B)$ 結束 ### 把 $\Sigma$ 零的部分都蓋掉  - $U$ 和 $V^T$ 對應的部分也蓋掉,仍然會等於 $A$ - 證明寫黑板,有點懶得自己證 ### 如果把 $\Sigma_k$ 的部分也蓋掉呢? 乘出來會是 $A'$ 不等於 $A$ - 但是 **$A'$ is the rank k-1 matrix minimizing $\|A-A'\|$** - 這裡的 $\|A-A'\|$ 就是用 Frobenius inner product 算出的 Frobenius norm - 沒有證明 ## SVD 可以對 data 做壓縮 [Low rank approximation using the singular value decomposition](https://www.youtube.com/watch?v=pAiVb7gWUrM) - 自己調整 k' 決定壓縮率 {%youtube pAiVb7gWUrM %} ## SVD 的歌 o_o [The SIngular Value Decomposition (SVD) song: It Had To Be U](https://www.youtube.com/watch?v=fKVRSbFKnEw) {%youtube fKVRSbFKnEw %}
×
Sign in
Email
Password
Forgot password
or
By clicking below, you agree to our
terms of service
.
Sign in via Facebook
Sign in via Twitter
Sign in via GitHub
Sign in via Dropbox
Sign in with Wallet
Wallet (
)
Connect another wallet
New to HackMD?
Sign up