Table 2 Load balance statistics about training and searching time for 64 and 128 nodes
From: Parallel multiple instance learning for extremely large histopathology image analysis
Stage | Comm. & Sync. | Computation | ||||
|---|---|---|---|---|---|---|
avg.[s] | max.[s] | avg./max. | avg.[s] | max.[s] | avg./max. | |
64-training | 58.1 | 64.1 | 0.91 | 254.5 | 310.0 | 0.82 |
64-searching | 275.4 | 294.4 | 0.94 | 363.7 | 589.9 | 0.62 |
128-training | 81.4 | 90.9 | 0.90 | 121.4 | 200.2 | 0.61 |
128-searching | 147.3 | 160.6 | 0.92 | 178.8 | 324.8 | 0.55 |