Skip to content

Commit

Permalink
updates V3
Browse files Browse the repository at this point in the history
  • Loading branch information
W Potosnak committed Apr 21, 2022
1 parent 4ff3e69 commit c6b6a72
Show file tree
Hide file tree
Showing 12 changed files with 479 additions and 2,529 deletions.
107 changes: 67 additions & 40 deletions auton_survival/estimators.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -76,6 +76,13 @@ def _fit_dcm(features, outcomes, vsize, val_data, random_seed, **hyperparams):
and columns as covariates.
outcomes : pd.DataFrame
A pandas dataframe with columns 'time' and 'event'.
vsize : float, default=0.15
Amount of data to set aside as the validation set.
Not applicable to 'rsf' and 'cph' models.
val_data : tuple
A tuple of the validation dataset features and outcomes of
'time' and 'event'.
If passed, vsize is ignored.
random_seed : int
Controls the rproduecibility of fitted estimators.
hyperparams : Optional arguments
Expand All @@ -100,10 +107,10 @@ def _fit_dcm(features, outcomes, vsize, val_data, random_seed, **hyperparams):

from .models.dcm import DeepCoxMixtures

k = hyperparams.get("k", 3)
k = hyperparams.get("k", 3)
layers = hyperparams.get("layers", [100])
batch_size = hyperparams.get("batch_size", 128)
learning_rate = hyperparams.get("learning_rate", 1e-3)
bs = hyperparams.get("batch_size", 128)
lr = hyperparams.get("learning_rate", 1e-3)
epochs = hyperparams.get("epochs", 50)
smoothing_factor = hyperparams.get("smoothing_factor", 1e-4)
gamma = hyperparams.get("gamma", 10)
Expand All @@ -113,9 +120,9 @@ def _fit_dcm(features, outcomes, vsize, val_data, random_seed, **hyperparams):
gamma=gamma,
smoothing_factor=smoothing_factor,
random_seed=random_seed)
model.fit(x=features, t=outcomes.time, e=outcomes.event, vsize=vsize,
val_data=val_data, iters=epochs, batch_size=batch_size,
learning_rate=learning_rate)
model.fit(x=features, t=outcomes.time, e=outcomes.event, vsize=vsize,
val_data=val_data, iters=epochs, batch_size=bs,
learning_rate=lr)

return model

Expand All @@ -139,6 +146,13 @@ def _fit_dcph(features, outcomes, vsize, val_data, random_seed, **hyperparams):
and columns as covariates.
outcomes : pd.DataFrame
A pandas dataframe with columns 'time' and 'event'.
vsize : float, default=0.15
Amount of data to set aside as the validation set.
Not applicable to 'rsf' and 'cph' models.
val_data : tuple
A tuple of the validation dataset features and outcomes of
'time' and 'event'.
If passed, vsize is ignored.
random_seed : int
Controls the reproducibility of called functions.
hyperparams : Optional arguments
Expand All @@ -147,7 +161,7 @@ def _fit_dcph(features, outcomes, vsize, val_data, random_seed, **hyperparams):
A list consisting of the number of neurons in each hidden layer.
- 'learning rate' : float, default=1e-3
Learning rate for the 'Adam' optimizer.
- 'bs' : int, default=100
- 'batch_size' : int, default=100
Learning is performed on mini-batches of input data.
This parameter specifies the size of each mini-batch.
- 'epochs' : int, default=50
Expand All @@ -161,15 +175,15 @@ def _fit_dcph(features, outcomes, vsize, val_data, random_seed, **hyperparams):
from .models.cph import DeepCoxPH

layers = hyperparams.get("layers", [100])
learning_rate = hyperparams.get("learning_rate", 1e-3)
bs = hyperparams.get("bs", 100)
lr = hyperparams.get("learning_rate", 1e-3)
bs = hyperparams.get("batch_size", 128)
epochs = hyperparams.get("epochs", 50)

model = DeepCoxPH(layers=layers, random_seed=random_seed)

model.fit(x=features, t=outcomes.time, e=outcomes.event, vsize=vsize,
val_data=val_data, iters=epochs, batch_size=batch_size,
learning_rate=learning_rate)
model.fit(x=features, t=outcomes.time, e=outcomes.event, vsize=vsize,
val_data=val_data, iters=epochs, batch_size=bs,
learning_rate=lr)

return model

Expand Down Expand Up @@ -271,7 +285,7 @@ def _fit_rsf(features, outcomes, random_seed, **hyperparams):
Parameters
-----------
features : pd.DataFrame
A pandas dataframe with rows corresponding to individual samples and
A pandas dataframe with rows corresponding to individual samples and
columns as covariates.
outcomes : pd.DataFrame
A pandas dataframe with columns 'time' and 'event'.
Expand Down Expand Up @@ -334,6 +348,13 @@ def _fit_dsm(features, outcomes, vsize, val_data, random_seed, **hyperparams):
columns as covariates.
outcomes : pd.DataFrame
A pandas dataframe with columns 'time' and 'event'.
vsize : float, default=0.15
Amount of data to set aside as the validation set.
Not applicable to 'rsf' and 'cph' models.
val_data : tuple
A tuple of the validation dataset features and outcomes of
'time' and 'event'.
If passed, vsize is ignored.
random_seed : int
Controls the reproducibility of called functions.
hyperparams : Optional arguments
Expand All @@ -348,7 +369,7 @@ def _fit_dsm(features, outcomes, vsize, val_data, random_seed, **hyperparams):
- `batch_size` : int, default=100
Learning is performed on mini-batches of input data. This parameter
specifies the size of each mini-batch.
- `lr` : float, default=1e-3
- `learning_rate` : float, default=1e-3
Learning rate for the 'Adam' optimizer.
- `epochs` : int, default=1
Number of complete passes through the training data.
Expand All @@ -366,15 +387,15 @@ def _fit_dsm(features, outcomes, vsize, val_data, random_seed, **hyperparams):
epochs = hyperparams.get("iters", 10)
distribution = hyperparams.get("distribution", "Weibull")
temperature = hyperparams.get("temperature", 1.0)
lr = hyperparams.get("lr", 1e-3)
bs = hyperparams.get("batch_size", 1.0)
lr = hyperparams.get("learning_rate", 1e-3)
bs = hyperparams.get("batch_size", 128)

model = DeepSurvivalMachines(k=k, layers=layers,
distribution=distribution,
temp=temperature,
random_seed=random_seed)

model.fit(x=features, t=outcomes.time, e=outcomes.event, vsize=vsize,
model.fit(x=features, t=outcomes.time, e=outcomes.event, vsize=vsize,
val_data=val_data, iters=epochs, learning_rate=lr, batch_size=bs)

return model
Expand Down Expand Up @@ -538,8 +559,8 @@ def __init__(self, model, random_seed=0, **hyperparams):
self.random_seed = random_seed
self.fitted = False

def fit(self, features, outcomes, vsize=None, val_data=None,
weights_train=None, weights_val=None, resample_size=1.0):
def fit(self, features, outcomes, vsize=0.15, val_data=None,
weights=None, weights_val=None, resample_size=1.0):

"""This method is used to train an instance of the survival model.
Expand All @@ -550,17 +571,19 @@ def fit(self, features, outcomes, vsize=None, val_data=None,
columns as covariates.
outcomes : pd.DataFrame
a pandas dataframe with columns 'time' and 'event'.
vsize : float
vsize : float, default=0.15
Amount of data to set aside as the validation set.
Not applicable to 'rsf' and 'cph' models.
val_data : tuple
A tuple of the validation dataset.
If passed vsize is ignored.
A tuple of the validation dataset features and outcomes of 'time'
and 'event'.
If passed, vsize is ignored.
Not applicable to 'rsf' and 'cph' models.
weights_train : list or np.array
a list or numpy array of importance weights for each sample.
weights_val : list or np.array
a list or numpy array of importance weights for each validation set sample.
a list or numpy array of importance weights for each validation
set sample.
Ignored if val_data is None.
resample_size : float
a float between 0 and 1 that controls the size of the resampled dataset.
Expand All @@ -571,25 +594,26 @@ def fit(self, features, outcomes, vsize=None, val_data=None,
Trained instance of a survival model.
"""

if (self.model=='cph') | (self.model=='rsf'):
if (vsize is not None) | (val_data is not None):
raise Exception("'vsize' and 'val_data' should be None for 'cph' and 'rsf' models.")

if weights_train is not None:
assert len(weights_train) == features.shape[0], "Size of passed weights \
if weights is not None:
assert len(weights) == features.shape[0], "Size of passed weights \
must match size of training data."
assert (weights_train>0.).any(), "All weights must be positive."
assert (vsize is not None) | (val_data is not None), "'vsize' or 'val_data' must \
be specified if weights are used."

weights = pd.Series(weights, index=data.index)
assert (weights>0.).any(), "All weights must be positive."

data = features.join(outcomes)

weights = pd.Series(weights, index=data.index)

if val_data is not None:
assert weights_val is not None, "Validation set weights must be \
specified."
assert len(weights_val) == val_data[0].shape[0], "Size of passed \
weights_val must match size of validation data."
assert (weights_val>0.).any(), "All weights_val must be positive."

data_train = data
data_val = val_data
data_val = val_data[0].join(val_data[1])
weights_train = weights

else:
data_train = data.sample(frac=1-vsize, random_state=self.random_seed)
data_val = data[~data.index.isin(data_train.index)]
Expand All @@ -605,12 +629,15 @@ def fit(self, features, outcomes, vsize=None, val_data=None,
frac = resample_size,
replace = True,
random_state = self.random_seed)

features = data_train_resampled[features.columns]
outcomes = data_train_resampled[outcomes.columns]

val_data = (data_val_resampled[features.columns],
val_data = (data_val_resampled[features.columns],
data_val_resampled[outcomes.columns])

if val_data is not None:
val_data = (val_data[0], val_data[1].time, val_data[1].event)

if self.model == 'cph':
self._model = _fit_cph(features, outcomes,
Expand Down Expand Up @@ -638,7 +665,7 @@ def fit(self, features, outcomes, vsize=None, val_data=None,

else:
raise NotImplementedError()

self.fitted = True
return self

Expand All @@ -652,7 +679,7 @@ def predict_survival(self, features, times):
a pandas dataframe with rows corresponding to individual samples
and columns as covariates.
times : float or list
a float or list of the times at which to compute the survival
a float or list of the times at which to compute the survival
probability.
Returns
Expand Down
Loading

0 comments on commit c6b6a72

Please sign in to comment.