Måns Karlsson and Hampus Engsner: Neural Network basics for the mathematical statistician

Abstract: So called neural networks is a hot family of modern statistical models aimed at prediction and classification. This lecture will be a non-technical orientation in the theory, usage and implementation of this family of models. Starting from scratch, we define a basic neural network and proceed to the general architecture of convolutional neural networks (CNNs), which are famed for their usage in automated image recognition. Complementing the theory, we also show how to swiftly fit a CNN in e.g. Python and round of with some model diagnostics and examples of impact on human society.