import os, os.path
from time import time
import numarray
from numarray import random_array
import random
DSN = "dbname=test port = 5435"
# in order to always generate the same random sequence
random.seed(19)
random_array.seed(19, 20)
def flatten(l):
"""Flattens list of tuples l."""
return map(lambda x: x[0], l)
def fill_arrays(start, stop):
col_i = numarray.arange(start, stop, type=numarray.Int32)
if userandom:
col_j = random_array.uniform(0, nrows, shape=[stop-start])
else:
col_j = numarray.array(col_i, type=numarray.Float64)
return col_i, col_j
# Generator for ensure pytables benchmark compatibility
def int_generator(nrows):
step = 1000*100
j = 0
for i in xrange(nrows):
if i >= step*j:
stop = (j+1)*step
import os, os.path
from time import time
import numarray
from numarray import random_array
import random
DSN = "dbname=test port = 5435"
# in order to always generate the same random sequence
random.seed(19)
random_array.seed(19, 20)
def flatten(l):
"""Flattens list of tuples l."""
return map(lambda x: x[0], l)
def fill_arrays(start, stop):
col_i = numarray.arange(start, stop, type=numarray.Int32)
if userandom:
col_j = random_array.uniform(0, nrows, shape=[stop - start])
else:
col_j = numarray.array(col_i, type=numarray.Float64)
return col_i, col_j
# Generator for ensure pytables benchmark compatibility
def int_generator(nrows):
step = 1000 * 100
j = 0
import bisect # for appending elements to a sorted list
import logging
#Major Packages
import numarray as na
import numarray.mlab
from numarray.random_array import randint, seed
from numarray.ieeespecial import setnan, getnan
#Library Specific Modules
import graph
import delegate
import distributions
import potentials
import inference
seed()
#logging.basicConfig(level= logging.INFO)
class BVertex(graph.Vertex):
def __init__(self, name, discrete = True, nvalues = 2, observed = True):
'''
Name neen't be a string but must be hashable and immutable.
if discrete = True:
nvalues = number of possible values for variable contained in Vertex
if discrete = False:
nvalues is not relevant = 0
observed = True means that this node CAN be observed
'''
graph.Vertex.__init__(self, name)
self.distribution = None
if discrete:
def main(argv=None):
if argv is None:
argv = sys.argv
if len(argv)!=6:
print "Usage: %s <method> <frequency> <vmin> <vmax> <datafile>"%argv[0]
return 1
hybrid = "%s_hybrid"%argv[1]
simplex = "%s_simplex"%argv[1]
dfile = argv[5]
hybrid_out = "%s.hybrid_output"%dfile
simplex_out = "%s.simplex_output"%dfile
temp_in = "%s.temp_input"%dfile
frequency = argv[2]
vmin = argv[3]
vmax = argv[4]
os.system("%s %s %s %s %s 200 300 3 %s"%(hybrid,dfile,frequency,vmin,vmax,hybrid_out))
fin = open(hybrid_out,"r")
params = []
paramnames = []
v = []
tin = []
tfit = []
for l in fin:
tokens = l.split()
if tokens[0]=='#':
if tokens[1]!="Attained":
paramnames.append(tokens[1])
params.append(float(tokens[2]))
else:
v.append(float(tokens[0]))
tin.append(float(tokens[1]))
tfit.append(float(tokens[2]))
fin.close()
tres = [tfit[i]-tin[i] for i in xrange(len(tin))]
rms = reduce(lambda x,y: x+y*y, tres,0.0)
rms = math.sqrt(rms/len(tres))
prmstr = ' '.join([str(i) for i in params])
rnd.seed()
meanparams = [0.0]*len(params)
rmsparams = [0.0]*len(params)
niter = 100
for i in xrange(niter):
ret = 1
while(ret!=0):
rmsarr = rnd.normal(0.0,rms,len(tfit))
trand = [tfit[j]+rmsarr[j] for j in xrange(len(tfit))]
fout = open(temp_in,"w")
for j in range(len(v)):
print >>fout,"%f\t%g"%(v[j],trand[j])
fout.close()
ret=os.system("%s %s %s %s %s %s %s"%(simplex,temp_in,frequency,vmin,vmax,prmstr,simplex_out))
simplex_params = []
fin = open(simplex_out,"r")
for l in fin:
tokens = l.split()
if tokens[0]=='#':
if tokens[1]!="Attained":
simplex_params.append(float(tokens[2]))
fin.close()
for j in xrange(len(params)):
meanparams[j]+=simplex_params[j]
rmsparams[j]+=simplex_params[j]*simplex_params[j]
rmsparams=map(lambda x,y: math.sqrt((x-(y**2)/niter)/niter),rmsparams,meanparams)
meanparams=[x/niter for x in meanparams]
print "Residual RMS: ",rms
print "Parameter\tMean\t\tSigma\t\tBest"
for i in range(len(paramnames)):
print "%s\t\t%f\t%f\t%f"%(paramnames[i],meanparams[i],rmsparams[i],params[i])
os.remove(simplex_out)
os.remove(temp_in)
return 0
