def impose_collapse(pairs, samples, weights):
"""collapse the weight and position of each pair (i,j) in pairs
Collapse is defined as weight[j] += weight[i] and weights[i] = 0,
with samples[j] = samples[i].
Inputs:
samples -- a list of sample points
weights -- a list of sample weights
pairs -- set of tuples of indices (i,j) where collapse occurs
For example:
>>> impose_collapse({(0,1),(0,2)},[1,2,3,4,5],[.2,.2,.2,.2,.2])
([1.5999999999999996, 1.5999999999999996, 1.5999999999999996, 4.5999999999999996, 5.5999999999999996], [0.6000000000000001, 0.0, 0.0, 0.2, 0.2])
>>> impose_collapse({(0,1),(3,4)},[1,2,3,4,5],[.2,.2,.2,.2,.2])
([1.3999999999999999, 1.3999999999999999, 3.3999999999999999, 4.4000000000000004, 4.4000000000000004], [0.4, 0.0, 0.2, 0.4, 0.0])
Note: is 'mean-preserving' for samples and 'norm-preserving' for weights
"""
samples, weights = list(samples), list(weights) # don't edit inputs
m = mean(samples, weights)
# allow negative indexing
pairs = zip(*tuple(
tuple(len(weights) + i if i < 0 else i for i in j)
for j in zip(*pairs)))
#XXX: any vectorized way to do this?
from mystic.tools import connected
for i, j in connected(pairs).iteritems():
v = weights[i]
for k in j:
v += weights[k]
weights[k] = type(v)(0.0)
samples[k] = samples[i]
weights[i] = v
return impose_mean(m, samples, weights), weights
def impose_collapse(pairs, samples, weights):
"""collapse the weight and position of each pair (i,j) in pairs
Collapse is defined as weight[j] += weight[i] and weights[i] = 0,
with samples[j] = samples[i].
Inputs:
samples -- a list of sample points
weights -- a list of sample weights
pairs -- set of tuples of indices (i,j) where collapse occurs
For example:
>>> impose_collapse({(0,1),(0,2)},[1,2,3,4,5],[.2,.2,.2,.2,.2])
([1.5999999999999996, 1.5999999999999996, 1.5999999999999996, 4.5999999999999996, 5.5999999999999996], [0.6000000000000001, 0.0, 0.0, 0.2, 0.2])
>>> impose_collapse({(0,1),(3,4)},[1,2,3,4,5],[.2,.2,.2,.2,.2])
([1.3999999999999999, 1.3999999999999999, 3.3999999999999999, 4.4000000000000004, 4.4000000000000004], [0.4, 0.0, 0.2, 0.4, 0.0])
Note: is 'mean-preserving' for samples and 'norm-preserving' for weights
"""
samples, weights = list(samples), list(weights) # don't edit inputs
m = mean(samples, weights)
# allow negative indexing
pairs = zip(*tuple(tuple(len(weights)+i if i<0 else i for i in j) for j in zip(*pairs)))
#XXX: any vectorized way to do this?
from mystic.tools import connected
for i,j in connected(pairs).iteritems():
v = weights[i]
for k in j:
v += weights[k]
weights[k] = type(v)(0.0)
samples[k] = samples[i]
weights[i] = v
return impose_mean(m, samples, weights), weights
def func(x, *args, **kwds):
x = copy.copy(x) #XXX: inefficient
for i,j in connected(mask).iteritems():
for k in j:
try: x[k] = x[i]
except IndexError: pass
pairs = list(mask) #XXX: inefficient
while pairs: # deal with the offset
indx,trac = zip(*pairs)
trac = set(trac)
for i in trac:
try: x[i] += offset
except IndexError: pass
indx = trac.intersection(indx)
pairs = [m for m in pairs if m[0] in indx]
return f(x, *args, **kwds)
def func(x, *args, **kwds):
x = copy.copy(x) #XXX: inefficient
for i,j in connected(mask).iteritems():
for k in j:
try: x[k] = x[i]
except IndexError: pass
pairs = list(mask) #XXX: inefficient
while pairs: # deal with the offset
indx,trac = zip(*pairs)
trac = set(trac)
for i in trac:
try: x[i] += offset
except IndexError: pass
indx = trac.intersection(indx)
pairs = [m for m in pairs if m[0] in indx]
return f(x, *args, **kwds)