https://github.com/camdavidsonpilon/lifelike

WIP predicted survival functions
https://github.com/camdavidsonpilon/lifelike

Last synced: about 1 year ago
JSON representation

WIP predicted survival functions

• Host: GitHub
• URL: https://github.com/camdavidsonpilon/lifelike
• Owner: CamDavidsonPilon
• License: mit
• Created: 2019-08-14T20:53:37.000Z (almost 7 years ago)
• Default Branch: master
• Last Pushed: 2019-08-29T19:30:40.000Z (almost 7 years ago)
• Last Synced: 2025-03-17T23:38:58.908Z (about 1 year ago)
• Language: Python
• Homepage:
• Size: 1.63 MB
• Stars: 37
• Watchers: 5
• Forks: 4
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          

## Lifelike

Simple neural network approach to predicting survival curves based on maximizing the likelihood. See introduction blog article [Non-parametric survival function prediction

](https://dataorigami.net/blogs/napkin-folding/non-parametric-survival-function-prediction).

```python

from jax.experimental.stax import Dense, Dropout, Tanh

from jax.experimental import optimizers

import lifelike.losses as losses

from lifelike import Model

from lifelike.callbacks import ModelCheckpoint, Logger

from lifelike.utils import dump, load

model = Model([

    Dense(20), Tanh,

    Dense(16), Tanh,

    Dropout(),

    Dense(10),

])

model.compile(optimizer=optimizers.adam,

              loss=losses.NonParametric(),

              weight_l2=0.1, smoothing_l2=10.0)

model.fit(x_train, t_train, e_train,

    epochs=1000,

    batch_size=32,

    callbacks=[ModelCheckpoint("filename.pickle"), Logger()]

)

model.predict(x_novel)

# serialization

dump(model, "filename.pickle")

model = load("filename.pickle")

model.fit(...)

```