def genlist(mini, maxi, step):
r = []
i = mini
while i < maxi:
r.append(i)
i += step
r.append(maxi)
return r
def accuracy_by_filter_score( data, title, sc=False ):
# Convert Groups of arrays into lists of scores by csore.
first, last, random = [], [], []
hmm, best, ind = [], [], []
sample_sizes = []
for j in xrange(0,995):
#trial = np.loadtxt( '../Data/Results/' + trial_name, delimiter = ',' )
#accuracies = means[::2]
#acc_std = stds[::2]
#softcalls = means[1::2]
#sc_std = stds[1::2]
trial = data[j]
accuracies = trial[::2]
softcalls = trial[1::2]
if sc:
f,l,r,h,b,i = softcalls
else:
f,l,r,h,b,i = accuracies
first.append( f ); last.append( l ); random.append( r )
hmm.append( h ) ; best.append( b ); ind.append( i )
#sample_sizes.append( trial_name.split('_')[3] )
x = cscores
#plt.plot( x, first, label='First', ls='--', c='k', lw=1)
#plt.plot( x, last, label='Last', ls='--', c='c', lw=1)
#plt.plot( x, random, label='Random', ls='--', c='m', lw=1 )
hmm, hmm_std = rolling_average(hmm)
best, best_std = rolling_average(best)
ind, ind_std = rolling_average(ind)
plt.plot( x, hmm, label='HMM')#, c='y', lw=1 )
plt.plot( x, best, label='Best')#, lw=1, c='r' )
plt.plot( x, ind, label='Ind')#, lw=1, c='b')
# plt.title( trial_name.split('_')[0] + ' - ' + trial_name.split('_')[1] \
# + ' - ' + trial_name.split('_')[2], fontsize=14 )
plt.title( title, fontsize=18 )
plt.xlabel( 'Read Cutoff', fontsize=14 )
plt.ylabel( 'Accuracy', fontsize=14 )
plt.xlim([0.0, .95])
plt.ylim( [0.5,0.85] )
plt.legend(loc=8, bbox_to_anchor=(0.5, 0.0),
ncol=2, fancybox=True, shadow=True)
plt.gca().invert_xaxis()
print 'plt.show'
plt.show()
plt.savefig( '/Users/Jvivian/Desktop/MR.svg', dpi=300)
def getListOfFirstWords(notes):
dd = {}
for nn in notes:
firstWord = nltk.pos_tag([nltk.word_tokenize(nn.contents)[0]])[0]
if firstWord[0] in dd:
dd[firstWord[0]][1] += 1
else:
dd[firstWord[0]] = [firstWord[1], 1]
rr = []
for key, pair in dd.items():
rr.append((key, pair[1]))
rr.sort(lambda x,y:cmp(x[1],y[1]))
return rr
def accuracy_by_filter_score( data, title, sc=False ):
# Convert Groups of arrays into lists of scores by csore.
first, last, random = [], [], []
hmm, best, ind = [], [], []
sample_sizes = []
for j in xrange(0,10):
#trial = np.loadtxt( '../Data/Results/' + trial_name, delimiter = ',' )
#accuracies = means[::2]
#acc_std = stds[::2]
#softcalls = means[1::2]
#sc_std = stds[1::2]
trial = data[j]
accuracies = trial[::2]
softcalls = trial[1::2]
if sc:
f,l,r,h,b,i = softcalls
else:
f,l,r,h,b,i = accuracies
first.append( f ); last.append( l ); random.append( r )
hmm.append( h ) ; best.append( b ); ind.append( i )
#sample_sizes.append( trial_name.split('_')[3] )
x = [ 0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1, 0.0 ]
plt.plot( x, first, label='first', ls='--', c='k')
plt.plot( x, last, label='last', ls='--', c='c')
plt.plot( x, random, label='random', ls='--', c='m' )
plt.plot( x, hmm, label='hmm', c='y' )
plt.plot( x, best, label='best', lw=2, c='r' )
plt.plot( x, ind, label='ind', lw=2, c='b')
# plt.title( trial_name.split('_')[0] + ' - ' + trial_name.split('_')[1] \
# + ' - ' + trial_name.split('_')[2], fontsize=14 )
plt.title( title )
plt.xlabel( 'Chunk Score Cutoff', fontsize=14 )
plt.ylabel( 'Accuracy', fontsize=14 )
plt.ylim( [0.5,1.0] )
plt.legend(loc='lower left', bbox_to_anchor=(1.02, 0), borderaxespad=0, fontsize=14)
plt.gca().invert_xaxis()
plt.show()
def rules(self, color):
win = []
block = []
open_four = []
block_open_four = []
random = []
for move in self.board.get_empty_points():
best = 5
for two_directions in [["N", "S"], ["NE", "SW"], ["E", "W"],
["SE", "NW"]]:
stats = self.line_rule(color, move, two_directions)
if stats["win"]:
best = 1
break
elif stats["block_win"]:
if best > 2:
best = 2
elif stats["open_four"]:
if best > 3:
best = 3
elif stats["block_open_four"]:
if best > 4:
best = 4
if best == 1:
win.append(move)
elif best == 2:
block.append(move)
elif best == 3:
open_four.append(move)
elif best == 4:
block_open_four.append(move)
else:
random.append(move)
if len(win) > 0:
return {"Win": win}
if len(block) > 0:
return {"BlockWin": block}
if len(open_four) > 0:
return {"OpenFour": open_four}
if len(block_open_four):
return {"BlockOpenFour": block_open_four}
if len(random) > 0:
return {"Random": random}
return "pass"
def createPopulation():
random = []
a = 0
while a < popSize:
# ini buat check banyak rule tiap chromosome. min 1 rule , max 3 rule
checker = numpy.random.randint(1, 5)
# bikin array data random, dengan isi 0/1
data = numpy.random.randint(2, size=popLength * checker)
arrcheck = numpy.array_split(data, checker)
valid = True
j = 0
# nge check kalo di dalam array random apakah salah 1 persyaratannya adalah kosong, ex: suhu = [0,0,0]
# kalo suhu = [0,0,0], maka chromosome nya salah dan akan di random lagi
while j < checker and valid:
# array suhu bit ke 1-3
suhu = arrcheck[j][0:3]
# array waktu bit ke 4-7
waktu = arrcheck[j][3:7]
#array cuaca bit ke 8-11
cuaca = arrcheck[j][7:11]
#array kelembapan bit ke 12-14
kelembapan = arrcheck[j][11:14]
# nge check kalo suhu nya [0,0,0]
if (not any(suhu)):
valid = False
# nge check kalo waktu nya [0,0,0,0]
elif (not any(waktu)):
valid = False
# nge check kalo cuaca nya [0,0,0,0]
elif (not any(cuaca)):
valid = False
# nge check kalo kelembapan nya [0,0,0]
elif (not any(kelembapan)):
valid = False
j = j + 1
# kalo data nya valid, maka di masukan ke dalam array, otherwise random ulang sampai data valid
if valid:
random.append(data)
a = a + 1
return random
def run(black, white, rep):
runs = np.zeros(black + white)
for i in range(rep):
if i % 1000 == 0:
print i
b = bag(black, white)
r = []
for i in range(black + white):
r.append(b.pull())
runs += np.array(r)
print np.mean(runs / rep)
print np.var(runs / rep)
plt.plot(np.zeros(len(runs)) + np.mean(runs / rep))
plt.plot(runs / rep)
plt.title("%d experiments with draws from a bag of %d white and %d black" %
(rep, white, black))
plt.legend(["mean number of whites", "whites pr. run"])
plt.xlabel("N")
plt.ylabel("P")
plt.show()
def showPlot3():
"""
Produces a plot comparing the two robot strategies.
"""
xvals = [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0]
standard = []
for i in range(len(xvals)):
time = runSimulation(1, 1.0, 20, 20, xvals[i], 100, StandardRobot)
standard.append(time)
random = []
for i in range(len(xvals)):
time = runSimulation(1, 1.0, 20, 20, xvals[i], 100, RandomWalkRobot)
random.append(time)
#print standard, random
pylab.plot(xvals, standard, 'r-', label="StandardRobot")
pylab.plot(xvals, random, 'b-', label="RandomWalkRobot")
pylab.xlabel('Fraction of Room to be Cleaned')
pylab.ylabel('Mean time')
pylab.legend(loc='upper left')
pylab.title(
'Comparison of StandardRobot and RandomWalkRobot in a 20x20 room')
pylab.show()
def ComponentQuery2(subj, pred, obj):
prefix = """
PREFIX rdf:<http://www.w3.org/2000/01/rdf-schema#>
PREFIX dbpd:<http://dbpedia.org/ontology/>
"""
r = []
if pred!=None and obj!=None:
qSelect_S = prefix + """
SELECT ?sub WHERE {
?sub """ + pred + """ """ + obj + """.
}"""
#r1 = m_graph.query(qSelect_S)
sparql.setQuery(qSelect_S)
r1 = sparql.query().convert()
if r1 != None: # may need one more variable to record source
r.append(r1)
if subj!=None and obj!=None:
qSelect_P = prefix + """
SELECT ?pred WHERE {
""" + subj + """ ?pred """ + obj + """.
}"""
#r2 = m_graph.query(qSelect_P)
sparql.setQuery(qSelect_P)
r2 = sparql.query().convert()
if r2 != None: # may need one more variable to record source
r.append(r2)
if subj!=None and pred!=None:
qSelect_O = prefix + """
SELECT ?obj WHERE {
""" + subj + """ """ + pred + """ ?obj.
}"""
#r3 = m_graph.query(qSelect_O)
sparql.setQuery(qSelect_O)
r3 = sparql.query().convert()
if r3 != None: # may need one more variable to record source
r.append(r3)
#if r!=[]:
# print(r)
#PrintQueryResult(r)
return r
with open('parted_idx.json','r') as f:
partitions = json.load(f)
with open('traj_viewpoint.json','r') as f:
traj_viewpoint = json.load(f)
#%%
part_2 = partitions['random'] + partitions['phrase']
phrase = []
random = []
for t_i in part_2:
ls = re.split('\, +|\. ',data[t_i[0]]['instructions'][t_i[1]])
ls = [elm for elm in ls if elm != '']
if len(ls) > 1 and len(phrase) < 7392:
phrase.append(t_i)
else:
random.append(t_i)
partitions['random'] = random
partitions['phrase'] = phrase
with open('parted_idx.json','w') as f:
json.dump(partitions,f)
#%% start constructing
hardNeg = []
gt = {}
#%% entity and random (which is reversal actually)
## load responses
with open('responses_train_new.json') as f:
resp_1 = json.load(f)
def heuristique_arrondi(H,data):
m = gurobipy.Model("MyModel")
V = len(H)
x = []
z = []
for i in range(V+1):
x2 = []
for j in range(V+1):
x2.append(m.addVar(vtype = gurobipy.GRB.CONTINUOUS, name = "x%d%d"%(i,j)))
x.append(x2)
for i in range(V+1):
z2 = []
for j in range(V+1):
z2.append(m.addVar(vtype = gurobipy.GRB.CONTINUOUS, name = "z%d%x"%(i,j)))
z.append(z2)
m.update()
obj = gurobipy.LinExpr();
obj = 0
#Fonction objectif
for i in range(V):
for j in range(V):
obj += evalCouple2(H[i],H[j],data) * x[i][j]
m.setObjective(obj,gurobipy.GRB.MAXIMIZE)
#1 arete sortante par sommet
for j in range(V+1):
m.addConstr(gurobipy.quicksum(x[i][j] for i in range(V+1))==1)
#1 arete entrante par sommet
for i in range(V+1):
m.addConstr(gurobipy.quicksum(x[i][j] for j in range(V+1))==1)
#pas d'arete sur un meme sommet
for i in range(V+1):
m.addConstr(x[i][i]==0)
#pas d'arete de a vers b et de b vers a simultanes
for i in range(V+1):
for j in range(V+1):
m.addConstr(x[i][j]+x[j][i]<=1)
#Contraintes de flots
#1
m.addConstr(gurobipy.quicksum(z[0][j] for j in range(1,V+1)) == V)
#2
for i in range(1,V+1):
m.addConstr( (gurobipy.quicksum(z[i][j] for j in range(1,V+1) if j != i)+1) == (gurobipy.quicksum(z[j][i] for j in range(V+1) if j!=i)) )
#3
for i in range(V+1):
list_j = [j for j in range(1,V+1) if j != i]
for j in list_j:
m.addConstr((z[i][j]+z[j][i]) <= (V)*(x[i][j]+x[j][i]))
#4
for i in range(V+1):
list_j = [j for j in range(1,V+1) if j != i]
for j in list_j:
m.addConstr(z[i][j]>=0)
#Resolution
m.optimize()
result = []
v = [ [x[i][j].x,i,j] for j in range(len(x[i])-1) for i in range(len(x)-1)]
v.sort(key=lambda y:y[0], reverse = True)
alreadyI = [0] * V
alreadyJ = [0] * V
compteur=0
while(compteur < V-1):
if(alreadyI[v[0][1]] == 0 and alreadyJ[v[0][2]]==0):
r = c.deepcopy(result)
r.append([[v[0][1]],[v[0][2]]])
r = deepTransfoPl(r)
b=True
for i in r :
if(i[0] == i[len(i)-1]):
b= False
if(b):
compteur = compteur+1
result = r
alreadyI[v[0][1]]=1
alreadyJ[v[0][2]]=1
del(v[0])
result = result[0]
result.insert(0,len(result))
return result
def createRandomChildren(n, num_variables):
random = []
for i in range(0, n):
random.append(np.array(np.random.choice(a=[-1, 1], size=(1, num_variables)))[0])
return random
