Skip to main content
. 2019 Nov 5;10:4646. doi: 10.1038/s41467-019-12552-4

Fig. 3.

Fig. 3

The Eighty Five Percent Rule applied to a multilayered neural network. Test accuracy vs training error rate on the MNIST dataset for the a Parity and b Magnitude tasks for 1000 different simulations. In both cases the test accuracy peaks at or near the optimal error rate. Each color corresponds to a different target training accuracy