Skip to main content
. Author manuscript; available in PMC: 2020 Sep 28.
Published in final edited form as: Conf Comput Vis Pattern Recognit Workshops. 2020 Jul 28;2020:3639–3648. doi: 10.1109/cvprw50498.2020.00425

Table 1.

Final train and test loss values after training the different Signal PI-Net and Image PI-Net models.

PI-Net Train Loss Test Loss
Signal PI-Net Time-steps = 250 0.00159 0.00158
Signal PI-Net Time-steps = 500 0.00187 0.00187
CIFAR10 Image PI-Net 1.99793 2.06193
CIFAR10 Image PI-Net FA 2.02095 2.04441
CIFAR10 Image PI-Net FS 0.51734 0.52560
SVHN Image PI-Net 1.53533 1.51732
SVHN Image PI-Net FA 1.57923 1.54195
SVHN Image PI-Net FS 0.41519 0.40955