TensorCox

Tensorflow implementation of Coxs' partial likelihood ¹ based on a counting process representation². Combining Cox's model with the counting process representation allows for more complex censoring patterns as well as the inclusion of time-varying covariates. Implementing this in Tensorflow with batch optimization methods gives us a powerful approach that can scale to big data problems and make full use of distributed computing environments. Since the algorithm is fully written in Tensorflow it is easy to integrate it into larger workflows and combine it with Neural Networks etc.

Installation

pip install TensorCox

Documentation

Documentation for TensorCox is available at https://tensorcox.readthedocs.io/en/latest/

References