Table 1 Summary of accuracy and timing results obtained by shallow learning processing of the original handwritten digits MNIST dataset82 and the Fashion-MNIST dataset83 using WETCOW inspired algorithm and test implementation based on the ideas of critically synchronized learning.
| Â | Accuracy | Time | Others |
|---|---|---|---|
MNISTDigits (without training) | 0.9858–0.9883 (117–142 errors per 10,000 samples) | Several seconds |  |
MNISTDigits (with training) | 0.9986 (14 errors per 10,000 samples) | Several minutes | 0.88–0.998 14 h for 0.9977 accuracy |
MNISTFashion (without training) | 0.9385 (615 errors per 10,000 samples) | Several seconds | Â |
MNISTFashion (with training) | 0.9742 (258 errors per 10,000 samples) | Several minutes | 0.444–0.897 From 1 to 50 h |
