Dropout: Interactive Demo

Dropout demo

Uncetainty in Neural Networks: Dropout- JS Demo

Demo of MC dropout in action over the same data as used in the Bayesian ensembling demo. Implementation is as originally found in this excellent blog about why uncertainty is important, with original code here, for the dropout demo here. We've edited colouring and data generation to match that used in the ensembling demo, but the core implementation of dropout is unchanged. We also added interactive ability to swap activation functions. The purpose of this demo was to allow some intuitive comparison with ensembles. Note that dropout struggles to capture interpolated uncertainty in between data points.

Browser not supported for Canvas. Get a real browser.

Hyperparameters

Activation function:
ReLU TanH

(Click 'Reset NNs' for new hyperparams to take affect)

Adapted from Yarin Gal https://github.com/yaringal/DropoutUncertaintyDemos, originally by Andrej Karpathy https://cs.stanford.edu/people/karpathy/convnetjs/demo1/regression.html.