Table 2.
Fraction of casualties with vital signs restored to the healthy target range based on the reinforcement learning (RL) model relative to the Vampire program.
| Available fluid units | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2 | 4 | 8 | 16 | 28 | 32 | 40 | 52 | 64 | 76 | 90 | ||
| Number of casualties | 32 |
1.18 (1.16) |
2.08 (0.75) |
2.39 (0.42) |
3.01 (1.07) |
3.26 (1.58) |
3.30 (1.64) |
3.84 (1.41) |
2.21 (0.43) |
1.23 (0.07) |
0.93 (0.14) |
* |
| 16 |
2.64 (0.74) |
3.23 (1.05) |
3.36 (1.55) |
3.36 (1.75) |
1.67 (0.19) |
1.20 (0.06) |
0.93 (0.14) |
* | * | * | * | |
| 8 |
2.99 (0.87) |
2.94 (1.04) |
3.13 (1.46) |
1.21 (0.08) |
0.93 (0.14) |
* | * | * | * | * | * | |
| 4 |
2.96 (1.09) |
3.18 (1.64) |
1.16 (0.09) |
0.93 (0.14) |
* | * | * | * | * | * | * | |
Fraction > 1.00 indicates that the RL model restored more casualties than the Vampire program. Resource-limited conditions are indicated in bold font determined based on the theoretical optimum, using the cardio-respiratory model. Data are presented as mean (standard deviation) of the ratios of the number of casualties restored by the RL model relative to those restored by the Vampire program for different numbers of casualties and available fluid units.
*Indicates that the values are equal to the value on their left.