Table 3:
MIMIC III results across frequent (S), few-shot (F), and zero-shot (Z) groups. We mark prior methods for MIMIC datasets that we implemented with a *.
| S | F | Z | Harmonic Average | |||||
|---|---|---|---|---|---|---|---|---|
| R@5 | R@10 | R@5 | R@10 | R@5 | R@10 | R@5 | R@10 | |
| Random | 0.000 | 0.000 | 0.000 | 0.000 | 0.038 | 0.052 | 0.000 | 0.000 |
| Logistic (Vani et al., 2017) * | 0.273 | 0.427 | 0.014 | 0.014 | - | - | - | - |
| CNN (Baumeletal., 2018)* | 0.269 | 0.413 | 0.058 | 0.085 | - | - | - | - |
| ACNN (Mullenbach et al., 2018) * | 0.288 | 0.458 | 0.130 | 0.168 | - | - | - | - |
| Match-CNN (Rios and Kavuluru, 2018) | 0.278 | 0.426 | 0.049 | 0.060 | - | - | - | - |
| ESZSL + W2V | 0.135 | 0.191 | 0.031 | 0.051 | 0.157 | 0.257 | 0.065 | 0.105 |
| ESZSL + W2V 2 | 0.127 | 0.189 | 0.031 | 0.048 | 0.148 | 0.305 | 0.063 | 0.102 |
| ESZSL + GRALS | 0.256 | 0.393 | 0.033 | 0.060 | 0.076 | 0.138 | 0.064 | 0.114 |
| ZACNN | 0.278 | 0.435 | 0.152 | 0.195 | 0.364 | 0.442 | 0.232 | 0.310 |
| ZAGCNN | 0.283 | 0.445 | 0.166 | 0.216 | 0.428 | 0.495 | 0.252 | 0.337 |