notebook.community

Edit and run



In [1]:

    
%load_ext autoreload
%autoreload 2
from xymodel import *



In [13]:

    
# MxM example at approx. critical temp
M = 16
X = classicalXY(16,1.1)



In [8]:

    
# Using standard IS with uniform proposal (does not appear in the paper)
samplesIS = 1000
X.importanceSampling(samplesIS)









    Out[8]:





528.31983966844894



In [25]:

    
# Using AIS with linear temperature scale (Matlab version used in paper)
samplesAIS = 100
beta = linspace(0,1,100)
X.annealedImportanceSampling(beta, samplesAIS)









    Out[25]:





656.68662711602155

The following code shows implementation examples of Random neighbour and Left-Right that appears in the paper "Sequential Monte Carlo for Graphical Models" by Naesseth et. al. (2014).



In [17]:

    
# Using random neighbour SMC for PGM
particles = 1000
index = arange(M**2)
random.shuffle(index)
X.smc(index,particles,NT=0.6,resamp='sys')









    Out[17]:





633.73322310708932



In [24]:

    
# Using left-right SMC for PGM
particles = 1000
index = np.arange(M**2)
X.smc(index,particles,NT=0.6,resamp='sys')









    Out[24]:





656.88175974001626



In [24]: