16 lines
351 B
Plaintext
16 lines
351 B
Plaintext
DATA GENERATION
|
|
|
|
1) construct N orthogonal curves of length T
|
|
e.g.: e1(t) = sqrt(2) * sin(pi*x)
|
|
e2(t) = sqrt(2) * cos(pi*x)
|
|
|
|
2) generate M x N scores:
|
|
for each of N curves
|
|
draw M scores from a Laplacian distribution
|
|
|
|
3) compose M data points
|
|
|
|
|
|
ALGORITHM
|
|
|
|
1) start with some random assignment of weights for a B-vector (B = 2) |