Loss Convolutional Layers Dense Units 0 0.2 0.1 0.08 0.06 0.04 0.02 0 0.1 0.2 0.2 Optimal 128 64 8 0.1 256 512 1 0.1 2 3 4 5 0.1 0.2 0.5 1 Loss (h) Loss (e) Loss (b) 50 0.5 0.5 0.5 1 1 1 128 64 8 0.1 256 512 0.2 0.2 0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 ReLU 0.1 ELU 100 Convolutional Filters Loss (i) Loss (f) Loss (c) 150 Trial (a) 0.5 0.5 0.5 1 1 1 200 32 16 0.1 64 0.2 0.2 0.1 0.08 0.06 0.04 0.02 0 0.1 128 0.2 1 0.1 2 3 4 5 250 Loss (j) Loss (g) Loss (d) 0.5 0.5 0.5 1 1 1 300 Figure 10. The hyperparameter optimization for the 3D edge. (a) The loss against trial number. (b)-(j) The hyperparameter dependence on the loss, for each of the parameters in Table 2. L2 Norm JUNE 2020 Activation Function * Dropout Dense Layers IEEE ROBOTICS & AUTOMATION MAGAZINE L1 Norm * Batch Size 76