By focusing on linear dimensionality reduction, we show how to visualize many dynamic phenomena in neural networks.