Hello, I am considering about increasing evaluation for deepsurv. I am wondering it's reasonable that use output risk $h_{\theta}(x)$ to calculate survival lines just imitating breslow and KM estimation method because they come from full likelihood of CoxPH. such the base line cumulative hazard function $$\Lambda0 (t) = \sum{i=1}^{N} \frac{\deltai}{ \sum{j \in R( T{(i)} ) } \exp(h{\theta}(x))} I( t \geq T_{(i)} )$$ and then, calculate timeROC. Thanks!
Hello, I am considering about increasing evaluation for deepsurv. I am wondering it's reasonable that use output risk $h_{\theta}(x)$ to calculate survival lines just imitating breslow and KM estimation method because they come from full likelihood of CoxPH. such the base line cumulative hazard function $$\Lambda0 (t) = \sum{i=1}^{N} \frac{\deltai}{ \sum{j \in R( T{(i)} ) } \exp(h{\theta}(x))} I( t \geq T_{(i)} )$$ and then, calculate timeROC. Thanks!