Table 3 The most successful configurations and success rates provided by the training algorithms used in the regression and classification models
From: A decision support system to determine optimal ventilator settings
| Â | Regression model | Classification model | |||||||
|---|---|---|---|---|---|---|---|---|---|
Learning algorithm | Hidden layer count | Process time (s) | FiO2 | Frequency | Tidal volume | Average success | Hidden layer count | Process time (s) | Pressure support/ Volume support |
Levenberg-Marquardtbackpropagation | 30 | 1,4 | 86,72% | 83,66% | 87,14% | 85,84% | 100 | 37 | 85,60% |
Bayesian regulationbackpropagation | 50 | 121 | 91,43% | 89,60% | 90,61% | 90,55% | 5 | 1,1 | 94,40% |
One step secantbackpropagation | 50 | 17 | 90,81% | 89,58% | 91,59% | 90,66% | 100 | 25,1 | 91,10% |
BFGS quasi-Newtonbackpropagation | 20 | 12 | 83,56% | 77,72% | 78,50% | 79,93% | 30 | 29,6 | 83,30% |
Cyclical order weight/biastraining | 5 | 126 | 67,41% | 73,46% | 74,31% | 71,73% | 15 | 179 | 92,20% |
Sequential order weight/bias training | 300 | 6 | 100,00% | 99,81% | 99,54% | 99,78% | 350 | 12 | 86,70% |