Problem Sheets - Artificial Intelligence in the Life Sciences - Ruhr University Bochum

Abstract

This record contains five problem sheets from the lecture Artificial Intelligence in the Life Sciences, held in the winter term 2025/26 at the Faculty of Biology and Biotechnology, Ruhr University Bochum. The exercises take students from classical machine learning to modern deep learning and explainable AI, and many of them use life-science examples such as microscopy, histopathology, genomics and medical imaging. Topics include: Sheet 1: k-means and single-linkage clustering, decision trees and Shannon entropy, bagging and random forests Sheet 2: component trees, Otsu thresholding, PCA, Kullback–Leibler divergence, t-SNE and convolution operations Sheet 3: image filters (Sobel, Gaussian), fully connected networks with forward and backward passes, U-Net and ResNet Sheet 4: dropout, self-attention, scaled dot-product and multi-head attention, positional encoding and transformers Sheet 5: saliency maps, class activation maps (CAM), attention-based interpretability in multiple instance learning (MIL), and Shapley values Each sheet is provided as LaTeX source (exam document class) with points for every subtask and a grading table. Sheets 4 and 5 require XeLaTeX.