We use Monte Carlo method to visualize the how a neural network allocates its decision space for different classes. The following figure is derived using the MNIST dataset.
Associate region of classes on MNIST dataset. Data are mapped into a 3D space using t-SNE.
