NEWS
survivalGPU 0.1.0 (2026-11-17)
- Initial CRAN submission.
coxphGPU(): Cox proportional hazards regression, GPU-accelerated,
with an interface modeled on survival::coxph().
wceGPU(): Weighted Cumulative Exposure (WCE) models, GPU-accelerated,
with an interface modeled on WCE::WCE().
bootstrap(): adds bootstrap confidence intervals to a model fitted by
coxphGPU() or wceGPU(), with patient-level resampling processed in
batches on the GPU. The model must be fitted with x = TRUE, plus
patient_id for coxphGPU(), so that the bootstrap resamples exactly the
data the model was fitted on.
wceGPU() accepts several candidate numbers of knots (e.g.
nknots = 1:3) and selects the best one by AIC or BIC; bootstrap()
repeats this selection in every replicate, so that the confidence
intervals account for the uncertainty on the number of knots.
use_cuda() to check whether CUDA drivers are detected; both models
fall back to CPU automatically when they are not.
- The formula, model frame and martingale residuals of
coxphGPU() are
computed by survival::coxph() itself: the package has no compiled
code.