x_Feed, H_Feed, y_Feed, noise_sigma2_Feed = gen_data.dataTrain(
0, params['SNR_dB_train']) # 固定信道
data_Feed = {
'x': x_Feed,
'y': y_Feed,
'H': H_Feed,
'noise_sigma2': noise_sigma2_Feed,
}
model_train(sess, MMNet_nodes, data_Feed)
if epoch % 100 == 0:
print('4', noise_sigma2_Feed)
print('5', model_est(sess, MMNet_nodes, data_Feed))
print('===========epoch%d,now_time is %s=========' %
(epoch + n * params['maxEpoch'],
datetime.datetime.now().strftime('%X')))
ser_MMNet, loss = model_loss(sess, MMNet_nodes, data_Feed)
print('ser_MMNet', ser_MMNet, 'loss=', loss, '\n')
loss_MMNet.append(loss)
loss_all['MMNet'] = loss_MMNet
train_ed = time.time()
print("MMNet Train time is: " + str(train_ed - train_st))
# Testing
results = {
'SNR_dBs':
np.arange(params['SNR_dB_min_test'], params['SNR_dB_max_test'],
params['SNR_step_test'])
}
ser_MMNets = model_eval(sess, params, MMNet_nodes, gen_data,
params['test_iterations'])
results['ser_MMNets'] = ser_MMNets
for n in range(params['nRounds']):
for epoch in range(params['maxEpoch']):
x_Feed, H_Feed, y_Feed, noise_sigma2_Feed = gen_data.dataTrain(
epoch, params['SNR_dB_train'])
data_Feed = {
'x': x_Feed,
'y': y_Feed,
'H': H_Feed,
'noise_sigma2': noise_sigma2_Feed,
}
model_train(sess, DetNet_nodes, data_Feed)
if epoch % 1000 == 0:
print('===========epoch%d,now_time is %s=========' %
(epoch + n * params['maxEpoch'],
datetime.datetime.now().strftime('%X')))
ser_DetNet, loss = model_loss(sess, DetNet_nodes, data_Feed)
print('ser_DetNet', ser_DetNet, 'loss=', loss, '\n')
loss_DetNet.append(loss)
loss_all['DetNet'] = loss_DetNet # 损失值
train_ed = time.time()
print("DetNet Train time is: " + str(train_ed - train_st))
if 'OAMPNet' in params['simulation_algorithms']:
loss_OAMPNet = []
print('===========OAMPNet is Training, now_time is %s=========' %
(datetime.datetime.now().strftime('%X')))
train_st = time.time()
for n in range(params['nRounds']):
for epoch in range(params['maxEpoch']):
x_Feed, H_Feed, y_Feed, noise_sigma2_Feed = gen_data.dataTrain(
epoch, params['SNR_dB_train'])
for epoch in range(params['maxEpoch']):
# x_Feed, H_Feed, y_Feed, noise_sigma2_Feed = gen_data.dataTrain(epoch, params['SNR_dB_train'])
x_Feed, H_Feed, y_Feed, noise_sigma2_Feed = gen_data.dataTrain(
epoch, params['SNR_dB_train']) # 固定信道
data_Feed = {
'x': x_Feed,
'y': y_Feed,
'H': H_Feed,
'noise_sigma2': noise_sigma2_Feed,
}
model_train(sess, DetNetSIC2_nodes, data_Feed)
if epoch % 100 == 0:
print('===========epoch%d,now_time is %s=========' %
(epoch + n * params['maxEpoch'],
datetime.datetime.now().strftime('%X')))
ser, loss = model_loss(sess, DetNetSIC2_nodes, data_Feed)
print('ser', ser, 'loss=', loss, '\n')
loss_all.append(loss)
train_ed = time.time()
print("Train time is: " + str(train_ed - train_st))
# Testing
results = {
'SNR_dBs':
np.arange(params['SNR_dB_min_test'], params['SNR_dB_max_test'],
params['SNR_step_test'])
}
sers = model_eval(sess, params, DetNetSIC2_nodes, gen_data,
params['test_iterations'])
results['ser_DetNetSIC2s'] = sers
print(sers)
print('=========== is Training, now_time is %s=========' % (datetime.datetime.now().strftime('%X')))
train_st = time.time()
for n in range(params['nRounds']):
for epoch in range(params['maxEpoch']):
x_Feed, H_Feed, y_Feed, noise_sigma2_Feed = gen_data.dataTrain(epoch, params['SNR_dB_train'])
data_Feed = {
'x': x_Feed,
'y': y_Feed,
'H': H_Feed,
'noise_sigma2': noise_sigma2_Feed,
}
model_train(sess, DetNetSIC3_node1, data_Feed)
if epoch % 100 == 0:
print('===========epoch%d,now_time is %s=========' % (epoch + n*params['maxEpoch'],
datetime.datetime.now().strftime('%X')))
ser, loss = model_loss(sess, DetNetSIC3_node1, data_Feed)
print('ser', ser, 'loss=', loss, '\n')
train_ed = time.time()
print("Train time is: "+str(train_ed-train_st))
train_st = time.time()
for n in range(params['nRounds']):
for epoch in range(params['maxEpoch']):
x_Feed, H_Feed, y_Feed, noise_sigma2_Feed = gen_data.dataTrain(epoch, params['SNR_dB_train'])
data_Feed = {
'x': x_Feed,
'y': y_Feed,
'H': H_Feed,
'noise_sigma2': noise_sigma2_Feed,
}
model_train(sess, DetNetSIC3_node2, data_Feed)