This is an implementation of "Overcoming catastrophic forgetting in neural networks" (https://arxiv.org/abs/1612.00796) for supervised learning in TensorFlow.

model.py defines a simple fully-connected network and methods to compute the diagonal of the Fisher information matrix.

experiment.ipynb trains and tests a single network on three MNIST classification tasks sequentially (i.e., once the network begins training on a given task, it is never exposed to previous task training data again).