shuffle=True,
drop_last=False)
dl_test = torch.utils.data.DataLoader(ds_test,
batch_size=x_validate.shape[0],
shuffle=False)
model = Net4(inputs, hiddenUnits)
# model = MLP4(inputs, hiddenUnits)
optimizer = torch.optim.Adam(model.parameters(),
lr=learningRate,
weight_decay=l2Penalty)
accuracy, loss, accuracy_validate, loss_validate = pre_train(model,
optimizer,
dl_train,
dl_test,
x_train,
y_train,
x_validate,
y_validate,
run,
outputDir,
maxEpoch=200)
accs.append(accuracy)
losses.append(loss)
accs_validate.append(accuracy_validate)
losses_validate.append(loss_validate)
dir = outputDir + 'pre_train_acc_loss_net4.npz'
np.savez(dir,
a=np.mean(accs, axis=0),
b=np.std(accs, axis=0),
c=np.mean(losses, axis=0),
d=np.std(losses, axis=0),
dl_test = torch.utils.data.DataLoader(ds_test,
batch_size=x_test.shape[0],
shuffle=False)
model = csnn_learnable_r(2, features, bias=BIAS)
if learnable_r:
model.set_lambda(LAMBDA)
model.set_miu(MIU)
optimizer = torch.optim.Adam(model.parameters(),
lr=learningRate,
weight_decay=l2Penalty)
accuracy, loss, accuracy_test, loss_test = pre_train(model,
optimizer,
dl_train,
dl_test,
x_train,
y_train,
x_test,
y_test,
run,
outputDir,
maxEpoch=50)
# accs.append(accuracy)
# losses.append(loss)
# accs_test.append(accuracy_test)
# losses_test.append(loss_test)
# dir = outputDir + 'pre_train_acc_loss_csnn.npz'
# np.savez(dir, a=np.mean(accs, axis=0), b=np.std(accs, axis=0),
# c=np.mean(losses, axis=0), d=np.std(losses, axis=0),
# e=np.mean(accs_test, axis=0), f=np.std(accs_test, axis=0),
# g=np.mean(losses_test, axis=0), h=np.std(losses_test, axis=0))
batch_size=batchSize,
shuffle=True,
drop_last=False)
dl_test = torch.utils.data.DataLoader(ds_test,
batch_size=x_validate.shape[0],
shuffle=False)
model = Net3(inputs, hiddenUnits)
optimizer = torch.optim.Adam(model.parameters(),
lr=learningRate,
weight_decay=l2Penalty)
pre_train(model,
optimizer,
dl_train,
dl_test,
x_train,
y_train,
x_validate,
y_validate,
run,
outputDir,
maxEpoch=10)
bestValidationAccs = []
AUCs = []
ACCs = []
epochs = 200
ALPHAs = None
for run in range(runs):
np.random.seed(seeds[run])
torch.manual_seed(seeds[run])
dl_train = torch.utils.data.DataLoader(ds_train,
np.random.seed(seed)
torch.manual_seed(seed)
for LAMBDA in LAMBDAS:
print('lambda {:.2f}'.format(LAMBDA))
for run in range(runs):
dl_train = torch.utils.data.DataLoader(ds_train, batch_size=batchSize, shuffle=True, drop_last=True)
# when batchNorm is involved in model, sample size in batch has to be > 1, so drop last just in case
dl_test = torch.utils.data.DataLoader(ds_test, batch_size=x_test.shape[0], shuffle=False)
model = Net3_learnable_r(inputs, hiddenUnits, bias=BIAS)
if learnable_r:
model.set_lambda(LAMBDA)
model.set_miu(MIU)
optimizer = torch.optim.Adam(model.parameters(), lr=learningRate,
weight_decay=l2Penalty)
pre_train(model, optimizer, dl_train, dl_test, x_train, y_train, x_test, y_test, run, outputDir, maxEpoch=1)
bestValidationAccs = []
AUCs = []
ACCs = []
epochs = 1000
ALPHAs = None
for run in range(runs):
dl_train = torch.utils.data.DataLoader(ds_train, batch_size=batchSize, shuffle=True, drop_last=True)
dl_test = torch.utils.data.DataLoader(ds_test, batch_size=x_test.shape[0], shuffle=False)
dl_combined = torch.utils.data.DataLoader(ds_combined, batch_size=x_combined.shape[0], shuffle=False)
PATH = outputDir + '/csnn_run{}_epoch{}.pth'.format(run, 0)
l = torch.load(PATH)
model = l['net']
optimizer = torch.optim.Adam(model.parameters(), lr=learningRate,
weight_decay=l2Penalty)