While it doesn't involve training, these 'confusion matrix' animations of NNs classifying images or digits are fun, too: http://ml4a.github.io/dev/demos/cifar_

While it doesn't involve training, these 'confusion matrix' animations of NNs classifying images or digits are fun, too: http://ml4a.github.io/dev/demos/cifar_confusion.html http://ml4a.github.io/dev/demos/cifar_confusion.html http://ml4a.github.io/dev/demos/mnist_confusion.html http://ml4a.github.io/dev/demos/mnist_confusion.html Something about the high-speed updating makes me think of WOPR, in 'War Games', scoring nuclear-war scenarios.