Table 2 Effects of downsampling on the performance
From: Task-driven knowledge graph filtering improves prioritizing drugs for repurposing
Dataset | Model | Original (MRR) | Modified (MRR) | %-Change |
|---|---|---|---|---|
Hetionet | TransE | 0.2232 | 0.2374 | \(+6.4\) |
DistMult | 0.2280 | 0.2517 | \(+9.3\) | |
ComplEx | 0.1975 | 0.2782 | \(+40.8\) | |
RESCAL | 0.2428 | 0.2940 | \(+21.1\) | |
ConvE | 0.1312 | 0.1647 | \(+25.5\) | |
Random | 0.0343 | 0.0331 | \(-3.5\) | |
DRKG | TransE | 0.0822 | 0.0938 | \(+14.2\) |
DistMult | 0.0899 | 0.0931 | \(+3.6\) | |
ComplEx | 0.0896 | 0.0950 | \(+5.9\) | |
RESCAL | 0.0578* | 0.0650* | \(+12.4\) | |
ConvE | 0.0618* | 0.0649* | \(+4.9\) | |
Random | 0.0102 | 0.0086 | \(-16.7\) |