def train_network(X, p, corr_g, data_type):
# Set the parameters for training deep knockoffs
pars = dict()
# Number of epochs
pars['epochs'] = 50
# Number of iterations over the full data per epoch
pars['epoch_length'] = 100
# Data type, either "continuous" or "binary"
pars['family'] = "continuous"
# Dimensions of the data
pars['p'] = p
pars['ncat'] = ncat
# List of categorical variables
pars['cat_var_idx'] = np.arange(0, (ncat * (num_cuts)))
# Number of discrete variables
pars['ncat'] = ncat
# Number of categories
pars['num_cuts'] = num_cuts
# Size of regularizer
# pars['regularizer'] = grid_results[0]
# Boolean for using different weighting structure for decorr
pars['use_weighting'] = False
# Multiplier for weighting discrete variables
pars['kappa'] = 50
# Size of the test set
pars['test_size'] = 0
# Batch size
pars['batch_size'] = int(0.5 * n)
# Learning rate
pars['lr'] = 0.01
# When to decrease learning rate (unused when equal to number of epochs)
pars['lr_milestones'] = [pars['epochs']]
# Width of the network (number of layers is fixed to 6)
pars['dim_h'] = int(10 * p)
# Penalty for the MMD distance
pars['GAMMA'] = training_params['GAMMA']
# Penalty encouraging second-order knockoffs
pars['LAMBDA'] = training_params['LAMBDA']
# Decorrelation penalty hyperparameter
pars['DELTA'] = training_params['DELTA']
# Target pairwise correlations between variables and knockoffs
pars['target_corr'] = corr_g
# Kernel widths for the MMD measure (uniform weights)
pars['alphas'] = [1., 2., 4., 8., 16., 32., 64., 128.]
# Save parameters
np.save('/artifacts/pars_' + data_type + '.npy', pars)
# Where to store the machine
checkpoint_name = "/artifacts/" + model + "_" + data_type
# Where to print progress information
logs_name = "/artifacts/" + model + "_progress.txt"
# Initialize the machine
machine = KnockoffMachine(pars,
checkpoint_name=checkpoint_name,
logs_name=logs_name)
# Train the machine
machine.train(X.values)
pars['dim_h'] = int(10*p)
# Penalty for the MMD distance
pars['GAMMA'] = training_params['GAMMA']
# Penalty encouraging second-order knockoffs
pars['LAMBDA'] = training_params['LAMBDA']
# Decorrelation penalty hyperparameter
pars['DELTA'] = training_params['DELTA']
# Target pairwise correlations between variables and knockoffs
pars['target_corr'] = corr_g
# Kernel widths for the MMD measure (uniform weights)
pars['alphas'] = [1., 2., 4., 8., 16., 32., 64., 128.]
# machine = KnockoffMachine(pars)
# machine.train(X_train.values)
# Save parameters
np.save('/artifacts/pars.npy', pars)
# Where to store the machine
checkpoint_name = "/artifacts/" + model
# Where to print progress information
logs_name = "/artifacts/" + model + "_progress.txt"
# Initialize the machine
machine = KnockoffMachine(pars, checkpoint_name=checkpoint_name, logs_name=logs_name)
# Train the machine
machine.train(X_train)
# Penalty for the MMD distance
pars['GAMMA'] = 1
# Penalty encouraging second-order knockoffs
pars['LAMBDA'] = 1
# Decorrelation penalty hyperparameter
pars['DELTA'] = 1
# Target pairwise correlations between variables and knockoffs
pars['target_corr'] = corr_g
# Kernel widths for the MMD measure (uniform weights)
pars['alphas'] = [1., 2., 4., 8., 16., 32., 64., 128.]
# Where the machine is stored
checkpoint_name = "../models/deepmodel"
# test to exclude 51
x_train = x_train.to_numpy()
# Initialize the machine
machine = KnockoffMachine(pars, checkpoint_name)
# Train the machine
#pdb.set_trace()
machine.train(x_train)
# Generate deep knockoffs
xk_train = machine.generate(x_train)
# Save knockoffs
pyreadr.write_rds("../data/derived_data/knockoffs.rds", pd.DataFrame(xk_train))
