questions regarding the loss function

[waystogetthere]

Hello Professor Ren,

Thanks for your contribution to the Survival Analysis community. In your paper DRSA(Deep Recurrent Survival Analysis), I found that the $\mathcal{L}_{uncensored}$ is confused:

Actually, the $t^i$ indicates the time when the event truly occur, why we should maximise the probability that the variable $z$ (denoting the time that the event truly occur) should be smaller than $t^i$? Or did I misunderstand something?

Thanks a lot!