-
Notifications
You must be signed in to change notification settings - Fork 5
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request #25 from nicolasaunai/mva
add minimum variance analysis
- Loading branch information
Showing
1 changed file
with
100 additions
and
0 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,100 @@ | ||
import numpy as np | ||
|
||
# L,M,N constants | ||
L = 2 | ||
M = 1 | ||
N = 0 | ||
|
||
|
||
class MVA(object): | ||
""" | ||
Minimum Variance Analysis | ||
""" | ||
|
||
def __init__(self, B): | ||
""" | ||
@param B : magnetic field | ||
""" | ||
|
||
self._B = B.copy() # magnetic field | ||
|
||
# not normalizing anymore... | ||
# B2 = B[:,0]**2 + B[:,1]**2 + B[:,2]**2 | ||
# self._B[:,0]/=np.sqrt(B2) | ||
# self._B[:,1]/=np.sqrt(B2) | ||
# self._B[:,2]/=np.sqrt(B2) | ||
|
||
self._LMN = np.ndarray(shape=(3, 3)) # LMN matrix | ||
self._eigen = None | ||
|
||
varM = np.ndarray(shape=(3, 3)) | ||
|
||
# build the variance matrix <BiBj> - <Bi><Bj> | ||
for i in np.arange(3): | ||
for j in np.arange(3): | ||
varM[i, j] = np.mean(self._B[:, i] * self._B[:, j]) - np.mean( | ||
self._B[:, i] | ||
) * np.mean(self._B[:, j]) | ||
|
||
# find the eigenvalues of varM and the associated eigenvectors | ||
w, v = np.linalg.eigh(varM) | ||
|
||
# now we want to sort the eigenvalues | ||
ind = np.argsort(w) # return the indices that would sort 'w' | ||
self._eigen = w[ind] | ||
|
||
vt = v.transpose() # eigh returns v[:,i] associated with w[i] | ||
# and we want the line v[i,:] with w[i] | ||
# so that vB is Blmn | ||
|
||
# then put the vectors in the same order as the eigenvalues | ||
for h in np.arange(3): | ||
self._LMN[h, :] = vt[ind[h], :] | ||
|
||
# legacy magnetopause : reverse N if pointing inside the magnetosphere | ||
# can be convention that Nx >0 | ||
if self._LMN[N, 0] < 0: | ||
self._LMN[N, :] *= -1 | ||
|
||
# same kind of convention we want Lz >0 | ||
if self._LMN[L, 2] < 0: | ||
self._LMN[L, :] *= -1 | ||
|
||
# M defined as cross product LxN | ||
self._LMN[1, 0] = ( | ||
self._LMN[0, 1] * self._LMN[2, 2] - self._LMN[0, 2] * self._LMN[2, 1] | ||
) | ||
self._LMN[1, 1] = ( | ||
self._LMN[2, 0] * self._LMN[0, 2] - self._LMN[2, 2] * self._LMN[0, 0] | ||
) | ||
self._LMN[1, 2] = ( | ||
self._LMN[0, 0] * self._LMN[2, 1] - self._LMN[0, 1] * self._LMN[2, 0] | ||
) | ||
|
||
# ========================================================== | ||
# ========================================================== | ||
def to_LMN(self, vector): | ||
""" | ||
transforms vector[:,3] into the LMN coord. syst. | ||
""" | ||
vlmn = np.ndarray(vector.shape) | ||
for i in np.arange(3): | ||
vlmn[:, i] = ( | ||
self._LMN[i, 0] * vector[:, 0] | ||
+ self._LMN[i, 1] * vector[:, 1] | ||
+ self._LMN[i, 2] * vector[:, 2] | ||
) | ||
|
||
return vlmn | ||
|
||
# ========================================================== | ||
|
||
# ========================================================== | ||
# ========================================================== | ||
def quality(self): | ||
""" | ||
returns the quality of MVA as the ratio intermediate/min eigenvalues | ||
""" | ||
return self._eigen[1] / self._eigen[0] | ||
|
||
# ========================================================== |