def train(data, mask):
print('---------------------------------------------')
print('Starting training')
v_input = tf.placeholder('float', [None, param_rbm['num_v_nodes']])
v_mask = tf.placeholder('float', [None, param_rbm['num_v_nodes']])
epoch = tf.placeholder('float', [1])
loss_op = rbm(v_input, v_mask, epoch)
init = tf.global_variables_initializer()
with tf.Session() as sess:
sess.run(init)
errsum = 0.0
for step in range(param_rbm['num_epochs']):
for batch in range(1, 2016):
start = (step * param_rbm['batch_size']) % 2015
end = min(start + param_rbm['batch_size'], 2015)
x = data[start:end]
m = mask[start:end]
loss = sess.run(loss_op, feed_dict={v_input: x, v_mask: m, epoch: [step]})
errsum = errsum + loss
print('epoch=%d, batch=%d, loss=%f' % (step, batch, loss))
print('epoch=%d, loss=%f' % (step, errsum / 2015))
print('Training Finished!')
print('---------------------------------------------')
cRBMHidden = tf.nn.sigmoid(tf.add(tf.matmul(content_x,cw),cb))
cinarybRBMSample = tf.nn.relu(tf.sign(cRBMHidden - tf.random_uniform(tf.shape(cRBMHidden))))
tRBMHidden = tf.nn.sigmoid(tf.add(tf.matmul(traffic_x,tw),tb))
tinarybRBMSample = tf.nn.relu(tf.sign(tRBMHidden - tf.random_uniform(tf.shape(tRBMHidden))))
RBMSample = 0
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
f2,f3 = sess.run([cinarybRBMSample,tinarybRBMSample], feed_dict={content_x: contentFeatures,traffic_x:trafficFeatures})
RBMSample = f2
RBMSample = np.hstack((RBMSample,f3))
BBRBM = rbm(70,50,learning_rate=0.001,momentum=0.8,rbm_type='bbrbm',relu_hidden = True,init_method= 'uniform')
BBRBM.plot=True
w2, b2, b3= BBRBM.pretrain(RBMSample,batch_size=100,n_epoches=100)
#no_GBRBM = rbm(122,80,learning_rate=0.001,momentum=0.8,rbm_type='gbrbm',relu_hidden = True)
#no_GBRBM.plot=True
#w, b1, b2= no_GBRBM.pretrain(allFeatures,batch_size=100,n_epoches=100)
#basic_GBRBM = rbm(90,70,learning_rate=0.001,momentum=0.8,rbm_type='gbrbm',relu_hidden = True)
#basic_GBRBM.plot=True
#w, b1, b2= basic_GBRBM.pretrain(basicFeatures,batch_size=100,n_epoches=100)
#content_GBRBM = rbm(13,10,learning_rate=0.001,momentum=0.8,rbm_type='gbrbm',relu_hidden = True)
#content_GBRBM.plot=True
wsavePath = 'E:/workspace for python/Experiment/multimodalExperiment/NSLKDD/RBMs/noModalityRBMPara/122_60_w.csv'
b1savePath = 'E:/workspace for python/Experiment/multimodalExperiment/NSLKDD/RBMs/noModalityRBMPara/122_60_b1.csv'
b2savePath = 'E:/workspace for python/Experiment/multimodalExperiment/NSLKDD/RBMs/noModalityRBMPara/122_60_b2.csv'
#def writeCsv(savePath,data):
# file = open(savePath, 'w', newline='')
# csvWriter = csv.writer(file)
# count = 0
# for row in data:
# temp_row=np.array(row)
# csvWriter.writerow(temp_row)
# count +=1
# file.close()
no_GBRBM = rbm(122,
60,
learning_rate=0.001,
momentum=0.8,
rbm_type='gbrbm',
relu_hidden=True)
no_GBRBM.plot = True
w, b1, b2 = no_GBRBM.pretrain(allFeatures, batch_size=100, n_epoches=100)
#basic_GBRBM = rbm(90,70,learning_rate=0.001,momentum=0.8,rbm_type='gbrbm',relu_hidden = True)
#basic_GBRBM.plot=True
#w, b1, b2= basic_GBRBM.pretrain(basicFeatures,batch_size=100,n_epoches=100)
#content_GBRBM = rbm(13,10,learning_rate=0.001,momentum=0.8,rbm_type='gbrbm',relu_hidden = True)
#content_GBRBM.plot=True
#w, b1, b2= content_GBRBM.pretrain(contentFeatures,batch_size=100,n_epoches=100)
#traffic_GBRBM = rbm(19,15,learning_rate=0.001,momentum=0.8,rbm_type='gbrbm',relu_hidden = True)
#traffic_GBRBM.plot=True