Table 1 Empirically selected hyperparameters in our auto-encoders
From: Estimation of gait normality index based on point clouds through deep auto-encoder
Training algorithm | RMSProp |
Loss function | MSE |
Initial learning rate | 0.0001 |
λ (L2-regularization) | 0.25 |
Momentum | 0.9 |
Batch size | 512 |
α (leaky ReLU) | 0.1 |
Number of epochs (without dropout) | 800 |
Number of epochs (with dropout) | 1600 |
Dropout probability | 0.3 |
Weight initialization | Xavier [10] |