Euclidean analog search
This module looks for analog patterns in a library dataset. Analogs are
selected according to a minimal Euclidean distance in the search space.
Two different search spaces have been implemented to date. There are
classes to search for analogs in a PCA truncated space and in a CCA
truncated one. Further details can be found in
Fernandez and J. Saenz
Analog search in CCA space.
(Submitted to Climate Research)
Contact the authors for a draft copy.
|
Imported modules
|
|
import Numeric
import math
import os
import pyclimate.bpcca
import pyclimate.mvarstatools
import pyclimate.pyclimateexcpt
import pyclimate.svdeofs
import pyclimate.tools
|
|
Functions
|
|
get_weights
|
|
|
get_weights
|
get_weights ( distarray, weightexp )
Returns the weight values for each analog
The weights are normalized so the sum of them all is 1. The
weights are inversely proportional to the weightexp power
of the distance.
Arguments:
-
distarray
- Array with the distances to the
smoothing (see the class
ANALOGSelector) nearest analogs
-
weightexp
- The exponent of the weights.
|
|
Classes
|
|
ANALOGSelector |
Reconstructs a field averaging over several analog patterns
|
CCAANALOG |
Analog search in the CCA space
|
EOFANALOG |
Analog search in the PCA space
|
__ANALOG |
Base class for the analog search. It cannot be instanciated!
|
|
|