Temporally-Consistent Survival Analysis Supplementary Material

This appendix is organized as follows. In Section A.1, we provide complete proofs for the results We develop a generalization of TCSR that considers multi-hop transitions in Section A.2. Finally, in Section A.3, we revisit connections with RL and sketch For convenience, we briefly recall each result before presenting a complete proof. Proposition 2. If A1-A2 are satisfied, then, for any initial row-stochastic matrix (k 1) Proposition 3. Algorithm 1 is equivalent the fixed-point iteration The last equality establishes the equivalence between (5) and Line 8 in Algorithm 1.A.2 TCSR( λ) We identify two cases of special interest. TCSR(0) is equivalent to Algorithm 1 presented in the main text. The survival function is closely related to fundamental objects in dynamic programming and reinforcement learning. The code is structured as follows.