Week 5 Flashcards
1
Q
PCA advantages
A
Efficient
Strong theoretical background
Can be used to store data efficiently (image compression)
Visual evaluation possible for small number of components
2
Q
Scaling matrix
A
3
Q
Rotation clockwise
A
4
Q
Rotation anticlockwise
A
5
Q
Reflection about x axis
A
6
Q
Reflection about y axis
A
7
Q
Reflection about origin
A
8
Q
Sheer in X direction
A
9
Q
Sheer in Y direction
A
10
Q
Sheer in x and y direction
A
11
Q
PCA algo
A
1) compute mean ‘row’ vector (1 x d)
2) compute mean row matrix (n x d)
3) B = X - X-
4) C = 1/n • BT • B
5) take eigenvectors for k largest eigenvalues
6) W = (v1,v2,…,vk)
7) multiply each data point
12
Q
How much variance did you capture in PCA
A
Where r = k
Closer to 1 means more variance captured
