Deep Learning

An example of how generative deep learning models can be used to transform one imaging modality into another. Taken and adapted from Irvine et al. (2025) https://arxiv.org/abs/2509.09235

Dive into the fascinating world of deep learning! In this course, you will explore the modern methods and architectures that drive scientific innovation in materials research today. From neurons and multilayer perceptrons to regularization & optimization, convolutional neural networks, recurrent neural networks, (variational) autoencoders, GANs, diffusion models, transfer learning, and even large language models – this course covers the full spectrum.

By the end of the course, you will
✨ be able to explain the fundamental methods and architectures of deep learning – directly linked to materials science and engineering,
✨ implement different networks and models in a Python-based deep learning framework such as PyTorch to solve specific domain-related questions,
✨ critically assess the model outputs and confidently interpret the results in the context of their scientific application.

Discover how deep learning is revolutionizing materials research – and take an active role in shaping that future! 

Learning goals

  • The students will be able to explain the fundamental methods and architectures in deep learning and their application for materials science and engineering.
  • They will be able to implement the different network and models in a Python-based deep learning framework (e.g. Pytorch) in order to solve a specific domain question.
  • Based on the model output, the students are able to critically assess and interpret the results in the context of the application.

Lecture content

  1. Neurons and multilayer perceptron
  2. Regularization and optimization
  3. Convolutional neural networks
  4. Recurrent neural networks
  5. (Variational) autoencoder
  6. Generative adversarial networks
  7. Diffusion models
  8. Transfer learning
  9. Large language models

Comments

  • The course will be taught in English.
  • All exercises will be carried out by programming using JupyterHubs.
  • The following prerequisites are recommended for the lecture: knowledge corresponding to the modules Mathematics for Engineers 1–3, Introduction to Programming, and Introduction to Data Science.
  • Completing a practical assignment, presenting it, and submitting a report are required pre-examination achievements.
  • The final examination is an oral exam.