SacCriticAggregate
Enums · AiDotNet.Enums
How a SAC policy update combines its two soft Q-functions.
Fields
| Field | Summary |
|---|---|
SacCriticAggregate FirstCritic |
The first Q-function alone, as the 2018a authors’ code (haarnoja/sac) computes the reparameterized loss. |
SacCriticAggregate Mean |
The average of the two, as softlearning (the 2018b authors’ code) computes the policy loss. |
SacCriticAggregate Minimum |
The smaller of the two, as both SAC papers specify for the policy gradient (2018a §5, 2018b §6). |
int value__ |