Table 1.
Classification error on the permutation invariant MNIST task (test set) obtained by averaging test errors of the last 5 epochs (for MNIST) and last epoch for EMNIST.
| Network | Classification error | |||||
|---|---|---|---|---|---|---|
| Dataset | eRBP+ (%) | eRBP× (%) | RBP (100) (%) | RBP (1) (%) | BP (100) (%) | BP (1) (%) |
| PI MNIST 784-100-10 | 3.77 (3.23) | 2.89 (2.81) | 2.74 (2.64) | 3.19 (2.98) | 2.25 (2.19) | 2.44 (2.39) |
| PI MNIST 784-200-10 | 3.53 (2.98) | 2.78 (2.53) | 2.13 (2.04) | 2.37 (2.33) | 1.85 (1.78) | 1.94 (1.88) |
| PI MNIST 784-500-10 | 2.86 (2.57) | 2.34 (2.23) | 2.00 (1.96) | 2.09 (2.06) | 1.63 (1.60) | 1.88 (1.80) |
| PI MNIST 784-200-200-10 | 2.96 (2.85) | 2.29 (2.22) | 2.50 (2.45) | 2.26 (2.25) | 1.80 (1.78) | 1.82 (1.74) |
| PI MNIST 784-500-500-10 | 2.36 (2.28) | 2.02 (1.96) | 2.24 (2.0) | 2.34 (2.31) | 1.90 (1.86) | 1.69 (1.56) |
| PI EMNIST 784-200-200-10 | 26.76 (25.26) | 21.83 (21.4) | 22.3 (20.18) | 32.37 (26.48) | 18.42 (16.06) | 18.23 (17.72) |
Numbers in parentheses indicates the value of nbatch. Peak accuracies were reported for GPU based simulations. Value between parentheses indicate the highest measured test accuracies.