-
Notifications
You must be signed in to change notification settings - Fork 214
/
survival_loss.py
167 lines (132 loc) · 6.47 KB
/
survival_loss.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
from abc import ABCMeta
import numpy as np
from sklearn.dummy import DummyRegressor
from sklearn.ensemble._gb_losses import RegressionLossFunction
from sklearn.utils.extmath import squared_norm
from ..nonparametric import ipc_weights
from ._coxph_loss import coxph_loss, coxph_negative_gradient
class DummySurvivalEstimator(DummyRegressor):
def __init__(self, strategy="mean", constant=None, quantile=None):
super().__init__(
strategy=strategy,
constant=constant,
quantile=quantile,
)
def fit(self, X, y, sample_weight=None):
_, time = y
return super().fit(X, time, sample_weight=sample_weight)
class SurvivalLossFunction(RegressionLossFunction, metaclass=ABCMeta): # noqa: B024
"""Base class for survival loss functions."""
# pylint: disable=abstract-method,no-self-use
def init_estimator(self):
return DummySurvivalEstimator(strategy='constant', constant=0.)
class CoxPH(SurvivalLossFunction):
"""Cox Partial Likelihood"""
# pylint: disable=no-self-use
def __call__(self, y, raw_predictions, sample_weight=None): # pylint: disable=unused-argument
"""Compute the partial likelihood of prediction ``y_pred`` and ``y``."""
# TODO add support for sample weights
return coxph_loss(y['event'].astype(np.uint8), y['time'], raw_predictions.ravel())
def negative_gradient(self, y, raw_predictions, sample_weight=None, **kwargs): # pylint: disable=unused-argument
"""Negative gradient of partial likelihood
Parameters
---------
y : tuple, len = 2
First element is boolean event indicator and second element survival/censoring time.
y_pred : np.ndarray, shape=(n,):
The predictions.
"""
ret = coxph_negative_gradient(
y['event'].astype(np.uint8), y['time'], raw_predictions.ravel())
if sample_weight is not None:
ret *= sample_weight
return ret
def update_terminal_regions(self, tree, X, y, residual, raw_predictions,
sample_weight, sample_mask,
learning_rate=0.1, k=0):
"""Least squares does not need to update terminal regions.
But it has to update the predictions.
"""
# update predictions
raw_predictions[:, k] += learning_rate * tree.predict(X).ravel()
def _update_terminal_region(self, tree, terminal_regions, leaf, X, y,
residual, raw_predictions, sample_weight):
"""Least squares does not need to update terminal regions"""
def _scale_raw_prediction(self, raw_predictions):
return raw_predictions
class CensoredSquaredLoss(SurvivalLossFunction):
"""Censoring-aware squared loss.
Censoring is taken into account by only considering the residuals
of samples that are not censored, or the predicted survival time
is before the time of censoring.
"""
# pylint: disable=no-self-use
def __call__(self, y, raw_predictions, sample_weight=None):
"""Compute the partial likelihood of prediction ``y_pred`` and ``y``."""
pred_time = y['time'] - raw_predictions.ravel()
mask = (pred_time > 0) | y['event']
return 0.5 * squared_norm(pred_time.compress(mask, axis=0))
def negative_gradient(self, y, raw_predictions, **kwargs): # pylint: disable=unused-argument
"""Negative gradient of partial likelihood
Parameters
---------
y : tuple, len = 2
First element is boolean event indicator and second element survival/censoring time.
y_pred : np.ndarray, shape=(n,):
The predictions.
"""
pred_time = y['time'] - raw_predictions.ravel()
mask = (pred_time > 0) | y['event']
ret = np.zeros(y['event'].shape[0])
ret[mask] = pred_time.compress(mask, axis=0)
return ret
def update_terminal_regions(self, tree, X, y, residual, raw_predictions,
sample_weight, sample_mask,
learning_rate=0.1, k=0):
"""Least squares does not need to update terminal regions.
But it has to update the predictions.
"""
# update predictions
raw_predictions[:, k] += learning_rate * tree.predict(X).ravel()
def _update_terminal_region(self, tree, terminal_regions, leaf, X, y,
residual, raw_predictions, sample_weight):
"""Least squares does not need to update terminal regions"""
def _scale_raw_prediction(self, raw_predictions):
np.exp(raw_predictions, out=raw_predictions)
return raw_predictions
class IPCWLeastSquaresError(SurvivalLossFunction):
"""Inverse probability of censoring weighted least squares error"""
# pylint: disable=no-self-use
def __call__(self, y, raw_predictions, sample_weight=None):
sample_weight = ipc_weights(y['event'], y['time'])
return (1.0 / sample_weight.sum()
* np.sum(sample_weight * ((y['time'] - raw_predictions.ravel()) ** 2.0)))
def negative_gradient(self, y, raw_predictions, **kwargs): # pylint: disable=unused-argument
return y['time'] - raw_predictions.ravel()
def update_terminal_regions(self, tree, X, y, residual, y_pred,
sample_weight, sample_mask,
learning_rate=0.1, k=0):
y_pred[:, k] += learning_rate * tree.predict(X).ravel()
def _update_terminal_region(self, tree, terminal_regions, leaf, X, y,
residual, pred, sample_weight): # pragma: no cover
pass
def _scale_raw_prediction(self, raw_predictions):
np.exp(raw_predictions, out=raw_predictions)
return raw_predictions
LOSS_FUNCTIONS = {
"coxph": CoxPH,
"squared": CensoredSquaredLoss,
"ipcwls": IPCWLeastSquaresError,
}