Theory and Modern Applications
From: Path classification by stochastic linear recurrent neural networks
Noise scale | Accuracy | Robustness test accuracies | Ratio | ||||||
---|---|---|---|---|---|---|---|---|---|
NRNN | min | max | avg. | NRNN | min | max | avg. | ||
1 | 100% | 97.37% | 100% | 99.21% | 100% | 94.74% | 100% | 99.21% | 100% |
1.5 | 97.37% | 86.84% | 100% | 95.00% | 97.37% | 92.11% | 100% | 95.79% | 100.83% |
2.5 | 97.37% | 65.79% | 86.84% | 75.26% | 97.37% | 76.32% | 94.74% | 83.95% | 111.54% |