Experiments modelling the EMNIST dataset on neural networks with varying widths and depths, Dropout layers, L1 & L2 Regularization, or Maxout Networks.
hwixley/EMNIST-NeuralNet-Regularisation-Experiments
A study of the problem of overfitting in deep neural networks, how it can be detected, and prevented using the EMNIST dataset. This was done by performing experiments with depth and width, dropout, L1 & L2 regularization, and Maxout networks.
Jupyter Notebook