|
|
|
| 1D clustering vs. |
||
| 2D "clustering" |
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| V = {Vge} = G x E expression
matrix (possibly a |
|||||||||||||||
| subset of all genes in the experiment) |
|||||||||||||||
| 1D
clustering: fit V ~ C X |
|||||||||||||||
| (C: each row contains one entry of 1 and the rest 0's; |
|||||||||||||||
| the rows of X are the cluster prototypes; relax to |
|||||||||||||||
| one nonzero entry in each row of C) |
|||||||||||||||
| PCA: the rows of X are orthogonal |
|||||||||||||||
| SCA: each row of C may contain up to p nonzero |
|||||||||||||||
| entries, where p is specified by the user |
|||||||||||||||