def main():
#nr=input('Enter the dimension N of the square lattice -->')
nr=4
reps=nr*nr # number of lattice sites
# generate complete dataset
(training_data,training_H,training_M)=mh.complete_dataset(reps)
train_labels=mh.gen_train_labels(training_data, training_H, training_M)
# compute frequesncies
counter=collections.Counter(train_labels)
val=counter.values()
freq=np.array(val)/np.float(len(train_labels))
#partition into training 90% and testing 10% datasets
#[train,test]=mh.partition(training_data,train_labels)
#train_set=train[0]
#train_y=train[1]
#test_set=test[0]
#test_y=test[1]
#test_classes=np.unique(test_y)
#train_classes=np.unique(train_y)
#training=[]
#testing=[]
#for i in range(len(train_set)):
#row=train_set[i]
#training.append(np.hstack((row,train_y[i])))
#for i in range(len(test_set)):
#row2=test_set[i]
#testing.append(np.hstack((row2,test_y[i])))
#np.reshape(testing,(len(testing),17))
#np.reshape(training,(len(training),17))
#train2=sorted(training,key=itemgetter(-1))
#test2=sorted(testing,key=itemgetter(-1))
train=[]
for i in range(len(training_data)):
row2=training_data[i]
train.append(np.hstack((row2,train_labels[i])))
np.reshape(train,(len(train),17))
train2=sorted(train,key=itemgetter(-1))
train_classes=np.unique(train_labels)
epochs=2
train_res=get_MH_sampled_IDs(train2,train_classes,freq,epochs)
for i in range(epochs):
res=train_res[i]
H=[]
for j in range(len(res)):
H.append(res[j][-2])
count=collections.Counter(H)
ens=[]
for k in np.sort(count.keys()):
ens.append(count[k])
freq2=np.array(ens)/np.float(len(H)) # estimated fdrequencies of the ditribution
freq1=freq2.tolist()
print "energy values:", np.sort(count.keys())
print "frequencies:",freq1
#histogr(H)
record(freq1)
def main():
#nr=input('Enter the dimension N of the square lattice -->')
nr=4
reps=nr*nr # number of lattice sites
# generate complete dataset
(training_data,training_H,training_M)=mh.complete_dataset(reps)
train_labels=mh.gen_train_labels(training_data, training_H, training_M)
# compute frequesncies
counter=collections.Counter(train_labels)
val=counter.values()
freq=np.array(val)/np.float(len(train_labels))
#partition into training 90% and testing 10% datasets
[train,test]=mh.partition(training_data,train_labels)
train_set=train[0]
train_y=train[1]
test_set=test[0]
test_y=test[1]
test_classes=np.unique(test_y)
train_classes=np.unique(train_y)
training=[]
testing=[]
for i in range(len(train_set)):
row=train_set[i]
training.append(np.hstack((row,train_y[i])))
for i in range(len(test_set)):
row2=test_set[i]
testing.append(np.hstack((row2,test_y[i])))
np.reshape(testing,(len(testing),17))
np.reshape(training,(len(training),17))
train2=sorted(training,key=itemgetter(-1))
test2=sorted(testing,key=itemgetter(-1))
print "Initial number of random seeds is set to 10"
print "It yields a training dataset of 10,000 samples per temperature"
print "The temperature range is set to [0.1 40.0] with step size 0.5"
expr=input("Would you like to increase the training dataset size? 'y'/'n'")
if expr=='y':
r=input("Enter a factor (integer) by which to increase the sample size")
elif expr=='n':
r=1
else:
r=1
train_res=get_MH_sampled_IDs(train2,train_classes,freq,r)
test_res=get_MH_sampled_IDs(test2,test_classes,freq,1)
write_to_csv(train_res, "training")
write_to_csv(test_res, "testing")
def main():
nr=4
reps=nr*nr # number of lattice sites
(training_data,training_H,training_M)=mh.complete_dataset(reps)
calc(training_data,nr)
output_filename='/home/nportman/UOIT_research/data/4by4.tar.gz'
source_dir='/home/nportman/UOIT_research/data/4by4'
make_tarfile(output_filename,source_dir)
def main():
nr=4
reps=nr*nr # number of lattice sites
(training_data,training_H,training_M)=mh.complete_dataset(reps)
classes=upsample(training_data,nr) # unique energy levels of the complete dataset
prop_train=0.8
(train,test)=partition(training_data,training_H,prop_train)
train_data=train[0]
test_data=test[0]
calc(train_data,classes,nr,1) # flag=1 for training
calc(test_data,classes,nr,2) # flag=2 for testing
def main():
# MH algorithm sampling IDs in the complete space
(complete, Hs, Ms)=complete_dataset(16)
prob=compute_boltz_weights(Hs)
print prob
train_res=mh.get_MH_samples(complete)
#train_labels=mh.gen_train_labels(complete, Hs, Ms)
#train=[]
#for i in range(len(complete)):
# row2=complete[i]
# train.append(np.hstack((row2,train_labels[i])))
#np.reshape(train,(len(train),17))
#train2=sorted(train,key=itemgetter(-1))
#train_classes=np.arange(0,len(levels),1)
#temp=40.0
#train_res=get_MH_sampled_IDs(train2,train_classes,temp)
H=[]
for i in range(len(train_res)):
H.append(train_res[i][-2])
histogr(H)
import gzip
import cPickle
import MH_dataset_generator as mh
import ising_model_princeton2 as I
import matplotlib.pyplot as plt
from sklearn.tree import DecisionTreeClassifier
from sklearn.decomposition import PCA
from sklearn.preprocessing import scale
from operator import itemgetter
nr=input('Enter the dimension N of the square lattice -->')
reps=nr*nr # number of lattice sites
# generate complete dataset
(training_data,training_H,training_M)=mh.complete_dataset(reps)
train_labels=mh.gen_train_labels(training_data, training_H, training_M)
#partition into training 90% and testing 10% datasets
[train,test]=mh.partition(training_data,train_labels)
train_set=train[0]
train_y=train[1]
test_set=test[0]
test_y=test[1]
training=[]
testing=[]
for i in range(len(train_set)):
row=train_set[i]
training.append(np.hstack((row,train_y[i])))
for i in range(len(test_set)):
row2=test_set[i]
testing.append(np.hstack((row2,test_y[i])))
def main():
nr = 4
reps = nr * nr # number of lattice sites
(training_data, training_H, training_M) = mh.complete_dataset(reps)
calc(training_data, nr)
import numpy as np
import random
import gzip
import cPickle
import MH_dataset_generator as mh
from sklearn.tree import DecisionTreeClassifier
from sklearn.decomposition import PCA
from sklearn.preprocessing import scale
nr=input('Enter the dimension N of the square lattice -->')
reps=nr*nr # number of lattice sites
# generate complete dataset
(training_data,training_H,training_M)=mh.complete_dataset(reps)
train_labels=mh.gen_train_labels(training_data, training_H, training_M)
[train,test]=mh.partition(training_data,train_labels)
train_set=train[0]
train_y=train[1]
test_set=test[0]
test_y=test[1]
train_data=mh.get_MH_samples(train_set)
print "... Computed training dataset"
test_data=mh.get_MH_samples(test_set)
print "... Computed testing dataset"
expr=input('Enter "H" for Hamiltonian energy estimation and "M" for magnetization -->')
if expr=="H":
train_labels=train_data[:,-2] # Hamiltonian
test_labels=test_data[:,-2]
