Table 3 Definition of different reward functions
From: Towards more efficient and robust evaluation of sepsis treatment with deep reinforcement learning
Indicator criterion | Rewards |
|---|---|
Sepsis3.0 | \(Reward_{3.0}=\sum _{i=0}^1W_itanh(S_i)\) |
\(Sepsis3.0^+\) | \(Reward_{3.0^+}=\sum _{i=2}^3W_itanh(S_i)\) |
Sepsis4.0 | \(Reward_{4.0}=\sum _{i=4}^5W_itanh(S_i)\) |
\(Sepsis(3.0+3.0^+)\) | \(Reward_{3.0+3.0^+}=reward_{3.0}+reward_{3.0^+}\) |
\(Sepsis(3.0+4.0)\) | \(Reward_{3.0+4.0}=reward_{3.0}+reward_{4.0}\) |
\(Sepsis(3.0^++4.0)\) | \(Reward_{3.0^++4.0}=reward_{3.0^+}+reward_{4.0}\) |
Sepsis(all) | \(Reward_{all}=reward_{3.0}+reward_{3.0^+}+reward_{4.0}\) |