def __init__(self, opt, logger):
super().__init__(opt, logger)
self.net_N = opt.net_N
self.x_dim = opt.x_dim
self.z_dim = opt.z_dim
self.n_classes = opt.n_classes
self.p_drop = opt.p_drop
self.modality = opt.modality
self.n_modality = len(self.modality)
self.clf_weights = opt.clf_weights
assert len(self.x_dim) == self.n_modality # check the length of x_dim
# init networks
self.net_q = [None] * self.n_modality
for i in range(self.n_modality):
self.net_q[i] = Q_net(self.net_N, self.x_dim[i], self.z_dim,
self.p_drop)
self.net_c = C_net(self.net_N, self.z_dim * self.n_modality,
self.n_classes, self.p_drop)
self._nets = self.net_q + [self.net_c]
# optimizers
self.optimizer_q = [None] * self.n_modality
for i in range(self.n_modality):
self.optimizer_q[i] = self.optimizer(self.net_q[i].parameters(),
lr=opt.lr,
**self.optimizer_params)
self.optimizer_c = self.optimizer(self.net_c.parameters(),
lr=opt.lr,
**self.optimizer_params)
self._optimizers = self.optimizer_q + [self.optimizer_c]
# scheduler
self.scheduler_q = [None] * self.n_modality
for i in range(self.n_modality):
self.scheduler_q[i] = self.scheduler(self.optimizer_q[i],
**self.scheduler_params)
self.scheduler_c = self.scheduler(self.optimizer_c,
**self.scheduler_params)
self._schedulers = self.scheduler_q + [self.scheduler_c]
# general
self.opt = opt
self._metrics = [
'loss_survival', 'c_index'
] # log the autoencoder loss and the classification loss
if opt.log_time:
self._metrics += ['t_survival']
# init variables
#self.input_names = ['features', 'labels']
self.init_vars(add_path=True)
# init weights
for i in range(self.n_modality):
self.init_weight(self.net_q[i])
self.init_weight(self.net_c)
def __init__(self, opt, logger):
super().__init__(opt, logger)
self.net_N = opt.net_N
self.x_dim = opt.x_dim
self.z_dim = opt.z_dim
self.n_classes = opt.n_classes
self.p_drop = opt.p_drop
self.modality = opt.modality
self.clf_weights = opt.clf_weights
# init networks
self.net_q = Q_net(self.net_N, self.x_dim, self.z_dim, self.p_drop)
self.net_p = P_net(self.net_N, self.x_dim, self.z_dim, self.p_drop)
self.net_c = C_net(self.net_N, self.z_dim, self.n_classes, self.p_drop)
self._nets = [self.net_q, self.net_p, self.net_c]
# optimizers
# self.optimizer and self.optimizer_params have been initialized in netinterface
self.optimizer_q = self.optimizer(self.net_q.parameters(),
lr=opt.lr,
**self.optimizer_params)
self.optimizer_p = self.optimizer(self.net_p.parameters(),
lr=opt.lr,
**self.optimizer_params)
self.optimizer_c = self.optimizer(self.net_c.parameters(),
lr=opt.lr,
**self.optimizer_params)
self._optimizers = [
self.optimizer_q, self.optimizer_p, self.optimizer_c
]
# schedulers
# self.scheduler and self.scheduler_params have been initialized in netinterface
self.scheduler_q = self.scheduler(self.optimizer_q,
**self.scheduler_params)
self.scheduler_p = self.scheduler(self.optimizer_p,
**self.scheduler_params)
self.scheduler_c = self.scheduler(self.optimizer_c,
**self.scheduler_params)
self._schedulers = [
self.scheduler_q, self.scheduler_p, self.scheduler_c
]
# general
self.opt = opt
self._metrics = [
'loss_mse', 'loss_survival', 'c_index'
] # log the autoencoder loss and the classification loss
if opt.log_time:
self._metrics += ['t_recon', 't_survival']
# init variables
self.init_vars(add_path=True)
# init weights
self.init_weight(self.net_q)
self.init_weight(self.net_p)
self.init_weight(self.net_c)
def __init__(self, opt, logger):
super().__init__(opt, logger)
self.net_N = opt.net_N
self.x_dim = opt.x_dim
self.z_dim = opt.z_dim
self.n_classes = opt.n_classes
self.p_drop = opt.p_drop
self.modality = opt.modality
self.n_modality = len(self.modality)
# init networks
self.net_q = Q_net(self.net_N, self.x_dim, self.z_dim, self.p_drop)
self.net_p = P_net(self.net_N, self.x_dim, self.z_dim, self.p_drop)
self.net_c = C_net(self.net_N, self.z_dim, self.n_classes, self.p_drop)
self._nets = [self.net_q, self.net_p, self.net_c]
# optimizers
self.optimizer_q = self.optimizer(self.net_q.parameters(),
lr=opt.lr,
**self.optimizer_params)
self.optimizer_p = self.optimizer(self.net_p.parameters(),
lr=opt.lr,
**self.optimizer_params)
self.optimizer_c = self.optimizer(self.net_c.parameters(),
lr=opt.lr,
**self.optimizer_params)
self._optimizers = [
self.optimizer_q, self.optimizer_p, self.optimizer_c
]
#schedulers
self.scheduler_q = self.scheduler(self.optimizer_q,
**self.scheduler_params)
self.scheduler_p = self.scheduler(self.optimizer_p,
**self.scheduler_params)
self.scheduler_c = self.scheduler(self.optimizer_c,
**self.scheduler_params)
self._schedulers = [
self.scheduler_q, self.scheduler_p, self.scheduler_c
]
# general
self.opt = opt
self._metrics = [
'loss_mse', 'loss_clf'
] # log the autoencoder loss and the classification loss
if opt.log_time:
self._metrics += ['t_recon', 't_clf']
# init variables
#self.input_names = ['features', 'labels']
self.init_vars(add_path=True)
# init weights
self.init_weight(self.net_q)
self.init_weight(self.net_p)
self.init_weight(self.net_c)
def __init__(self, opt, logger):
super().__init__(opt, logger)
self.net_N = opt.net_N
self.x_dim = opt.x_dim
self.z_dim = opt.z_dim
self.n_classes = opt.n_classes
self.p_drop = opt.p_drop
self.modality = opt.modality[
0] # get the fisrt element since changed this argument to list
self.clf_weights = opt.clf_weights
# init networks
self.net_q = Q_net(self.net_N, self.x_dim, self.z_dim, self.p_drop)
self.net_c = C_net(self.net_N, self.z_dim, self.n_classes, self.p_drop)
self._nets = [self.net_q, self.net_c]
# optimizers
# self.optimizer and self.optimizer_params have been initialized in netinterface
self.optimizer_q = self.optimizer(self.net_q.parameters(),
lr=opt.lr,
**self.optimizer_params)
self.optimizer_c = self.optimizer(self.net_c.parameters(),
lr=opt.lr,
**self.optimizer_params)
self._optimizers = [self.optimizer_q, self.optimizer_c]
# schedulers
# self.scheduler and self.scheduler_params have been initialized in netinterface
self.scheduler_q = self.scheduler(self.optimizer_q,
**self.scheduler_params)
self.scheduler_c = self.scheduler(self.optimizer_c,
**self.scheduler_params)
self._schedulers = [self.scheduler_q, self.scheduler_c]
# general
self.opt = opt
self._metrics = [
'loss_clf'
] # log the autoencoder loss and the classification loss
if opt.log_time:
self._metrics += ['t_clf']
# init variables
self.init_vars(add_path=True)
# init weights
self.init_weight(self.net_q)
self.init_weight(self.net_c)