This script is an exercise to apply neural networks to a classification problem using PyTorch. The dataset (GSE6008, taken from https://sbcb.inf.ufrgs.br/cumida) contains genetic markers (>20000) associated with particular types of ovarian cancer. The script applies PCA to reduce the number of features and then trains a neural network with two hidden layers in order to predict the cancer type. It uses tensorboard to log the losses during the training and then computes the accuracy of the network's productions on the test data.