tafsiri Refresh look and feel of examples: addition-rnn, cart-pole and mnist (#…

* Add shared css styles
* Replaces custom vis with tfjs-vis.
* update addition-rnn, cart-pole and mnist (layers)
Latest commit 5f7ac71 Nov 6, 2018


TensorFlow.js Example: Training MNIST

This example shows you how to train MNIST (using the layers API).

You can check out the tutorial that accompanies this example here.

This model will compute accuracy over 1000 random test set examples every 5 steps, plotting loss and accuracy as the model is training. Training time can be reduced by computing accuracy over fewer examples less often.

Note: currently the entire dataset of MNIST images is stored in a PNG image we have sprited, and the code in data.js is responsible for converting it into Tensors. This will become much simpler in the near future.

See this example live!