def build_mlps(kind, PV,true_out ,path, datasets, name):
conf = read_parser(path)
learning_rate = float( conf['parm']['learning_rate'])
L1_reg = float( conf['parm']['l1_reg'])
L2_reg = float( conf['parm']['l2_reg'])
n_epochs = int( conf['parm']['n_epochs'])
batch_size = int( conf['parm']['batch_size'])
pre_run = int( conf['parm']['pre_run'])
n_in = int( conf['input_layer']['input_num'])
n_hidden = []
h_activation = []
n_out = int( conf['out_layer']['out_num'])
for i in xrange(len(conf)):
tem = 'hidden_layer_' + str(i)
if tem in conf.keys():
n_hidden.append(int(conf[tem]['hidden_num']))
h_activation.append(int(conf[tem]['activation']))
out = test_mlp(pre_run = pre_run,
kind = kind, \
PV =PV, \
true_out = true_out, \
name = name, \
learning_rate=learning_rate, \
L1_reg=L1_reg, \
L2_reg=L2_reg, \
n_epochs=n_epochs, \
datasets=datasets, \
batch_size=batch_size, \
n_hidden=n_hidden, \
h_activation = h_activation, \
n_out=n_out)
return out
def build_mlps(kind, PV, true_out, path, datasets, name):
conf = read_parser(path)
learning_rate = float(conf['parm']['learning_rate'])
L1_reg = float(conf['parm']['l1_reg'])
L2_reg = float(conf['parm']['l2_reg'])
n_epochs = int(conf['parm']['n_epochs'])
batch_size = int(conf['parm']['batch_size'])
pre_run = int(conf['parm']['pre_run'])
n_in = int(conf['input_layer']['input_num'])
n_hidden = []
h_activation = []
n_out = int(conf['out_layer']['out_num'])
for i in xrange(len(conf)):
tem = 'hidden_layer_' + str(i)
if tem in conf.keys():
n_hidden.append(int(conf[tem]['hidden_num']))
h_activation.append(int(conf[tem]['activation']))
out = test_mlp(pre_run = pre_run,
kind = kind, \
PV =PV, \
true_out = true_out, \
name = name, \
learning_rate=learning_rate, \
L1_reg=L1_reg, \
L2_reg=L2_reg, \
n_epochs=n_epochs, \
datasets=datasets, \
batch_size=batch_size, \
n_hidden=n_hidden, \
h_activation = h_activation, \
n_out=n_out)
return out
def build_cnn(path):
conf = read_parser(path)
learning_rate = float(conf['parm']['learning_rate'])
n_epochs = int(conf['parm']['n_epochs'])
batch_size = int(conf['parm']['batch_size'])
input_layer = conf['input_layer']
ConvPool = {}
hidden = {}
lr_layer = conf['lr_layer']
for i in xrange(len(conf)):
tem = 'ConvPool' + str(i)
if tem in conf.keys():
ConvPool[tem] = conf[tem]
print ConvPool
for i in xrange(len(conf)):
tem = 'hidden_layer_' + str(i)
if tem in conf.keys():
hidden[tem] = conf[tem]
print hidden
"""
def build_cnn(path):
conf = read_parser(path)
learning_rate = float( conf['parm']['learning_rate'])
n_epochs = int( conf['parm']['n_epochs'])
batch_size = int( conf['parm']['batch_size'])
input_layer = conf['input_layer']
ConvPool = {}
hidden = {}
lr_layer = conf['lr_layer']
for i in xrange(len(conf)):
tem = 'ConvPool' + str(i)
if tem in conf.keys():
ConvPool[tem] = conf[tem]
print ConvPool
for i in xrange(len(conf)):
tem = 'hidden_layer_' + str(i)
if tem in conf.keys():
hidden[tem] = conf[tem]
print hidden
"""
def build_dbn(kind, PV,true_out ,path, datasets, name):
conf = read_parser(path)
finetune_lr = float( conf['parm']['finetune_lr'])
pretrain_lr = float( conf['parm']['pretrain_lr'])
pretraining_epochs = int( conf['parm']['pretraining_epochs'])
training_epochs = int( conf['parm']['training_epochs'])
batch_size = int( conf['parm']['batch_size'])
k = int( conf['parm']['k'])
pre_run = int( conf['parm']['pre_run'])
input_num = int( conf['input_layer']['input_num'])
n_hidden = []
h_activation = []
out_num = int( conf['out_layer']['out_num'])
for i in xrange(len(conf)):
tem = 'hidden_layer_' + str(i)
if tem in conf.keys():
n_hidden.append(int(conf[tem]['hidden_num']))
h_activation.append(int(conf[tem]['activation']))
print n_hidden
out = test_DBN(pre_run = pre_run,
kind = kind, \
PV =PV, \
true_out = true_out, \
path = path, \
k = k, \
name = name, \
finetune_lr=finetune_lr, \
pretraining_epochs=pretraining_epochs,\
training_epochs=training_epochs, \
pretrain_lr = pretrain_lr, \
datasets=datasets, \
batch_size=batch_size, \
hidden_layers_sizes=n_hidden , \
h_activation = h_activation , \
input_num=input_num, \
out_num=out_num)
return out
def build_cnn(kind, PV, true_out, path, datasets, name):
conf = read_parser(path)
learning_rate = float(conf['parm']['learning_rate'])
n_epochs = int(conf['parm']['n_epochs'])
batch_size = int(conf['parm']['batch_size'])
pre_run = int(conf['parm']['pre_run'])
input_layer = conf['input_layer']
ConvPool = {}
hidden = {}
out_layer = conf['out_layer']
for i in xrange(len(conf)):
tem = 'ConvPool' + str(i)
if tem in conf.keys():
ConvPool[tem] = conf[tem]
for i in xrange(len(conf)):
tem = 'hidden_layer_' + str(i)
if tem in conf.keys():
hidden[tem] = conf[tem]
out = cnn( pre_run = pre_run,
kind = kind, \
PV =PV, \
true_out = true_out, \
learning_rate = learning_rate, \
n_epochs = n_epochs, \
datasets =datasets, \
batch_size=batch_size, \
path = path , \
name = name, \
input_layer=input_layer, \
hidden=hidden, \
ConvPool=ConvPool, \
out_layer=out_layer)
return out
def build_cnn(kind, PV,true_out ,path, datasets, name):
conf = read_parser(path)
learning_rate = float( conf['parm']['learning_rate'])
n_epochs = int( conf['parm']['n_epochs'])
batch_size = int( conf['parm']['batch_size'])
pre_run = int( conf['parm']['pre_run'])
input_layer = conf['input_layer']
ConvPool = {}
hidden = {}
out_layer = conf['out_layer']
for i in xrange(len(conf)):
tem = 'ConvPool' + str(i)
if tem in conf.keys():
ConvPool[tem] = conf[tem]
for i in xrange(len(conf)):
tem = 'hidden_layer_' + str(i)
if tem in conf.keys():
hidden[tem] = conf[tem]
out = cnn( pre_run = pre_run,
kind = kind, \
PV =PV, \
true_out = true_out, \
learning_rate = learning_rate, \
n_epochs = n_epochs, \
datasets =datasets, \
batch_size=batch_size, \
path = path , \
name = name, \
input_layer=input_layer, \
hidden=hidden, \
ConvPool=ConvPool, \
out_layer=out_layer)
return out
