The project requires implementing 2 new objective loss functions - one for survival loss and another is for binomial loss. Survival loss includes accelerated failure time model with left, right and interval censored outputs and of time-to-event following gaussian and logistic distribution. The binomial loss includes constraints over a number of trials. This requires calculating gradient , hessian and loss metric for each loss functions and create new changes in data structure to implement it correctly.


Avinash Barnwal-1


  • Toby Hocking
  • Hyunsu Cho