Machine learning is becoming a transformational force in our society, and will profoundly impact humanity in ways both good and bad. On the good side, recent scientific breakthroughs such as having solved the protein folding problem will dramatically accelerate the development of new drugs and vaccines; on the bad side, we are facing a future in which deep fakes are used for manipulation, autonomous weapons roam our skies and machine learning supports surveillance in totalitarian states.
On the scientific side, machine learning is making its weight felt across all disciplines. Some go so far as to postulate a fifth paradigm of scientific discovery, fueled by machine learning.
Contents
Physics of Machine Learning: Highlight core physics concepts that drive ML
Machine Learning for Physics: Equip you with tools to help conduct, and interpret, future experiments
The course introduces some of the most important techniques for inference, and for regression, classification, dimension reduction and density estimation; and it emphasizes the physical ideas and laws needed to make these work. See below for a more detailed curriculum.Taking part, and admin stuff
- If, after browsing the FAQ below, you believe this course is for you: Then please register here.
- The course starts with a python refresher in the tutorial on Oct 17th or 18th (identical content). Unless you are familiar with python and it's basic scientific stack (jupyter, numpy, matplotlib, scipy), please take part to help you solve the computational exercises.
- The main lectures are on Tuesdays and Thursdays from 09h15-11h00 in Großer Hörsaal, Philosophenweg 12.
- Identical plenary tutorials are offered on Mondays and Tuesdays, from 16h15-18h00. Pick whichever day suits you best. Tutorials are held in KIP, INF 227 HS2.
FAQ
- Q: Do I need prior knowledge in machine learning?
A: No.
- Q: I just want to learn the basics. Is this the right course?
A: This course will have a steep learning curve; if you only want to cover the basics, you probably find easier alternatives.
- Q: Is this course about deep learning?
A: Neural networks will play an important role; but this course is more about principles. For sure we will not discuss details of the latest architectures.
- Q: Will this course be repeated next year?
A: Yes, like every MSc core course. In winter semester 2023, the course will be given by Jan Pawlowski and Tilman Plehn.
- Q: I do not find the course listed as MSc core course in the MSc module handbook. Why?
A: Because the version on the departmental's website has not been updated yet; this will happen before start of the teaching term.
- Q: Is there a text book?
A: The book with the biggest overlap is the soon-to-be-published Murphy, Probabilistic Machine Learning: Advanced Topics. Full pdf available here.
- Q: Exam modalities?
- A: To be admitted to the written exam at the end of the semester, you need to gain 50% of the points in the exercise sheets.
Preliminary curriculum
- Introduction & linear dimension reduction
- Nonlinear dimension reduction: connection to stat. mechanics
- Nonparametric density estimation
- Basic clustering techniques, review of information theory
- Classification, take 1: discriminative
- Review: Multivariate distributions, Bayes theorem, conjugate priors
- Classification, take 2: parametric / generative
- Regression
- Regularization, linear differential operators
- Gaussian processes
- Classification, take 3: logistic regression, generalized linear models
- Multi-layer perceptrons
- Multi-layer perceptrons: capacity
- Deep neural networks
- Architectures
- Directed Probabilistic Graphical Models
- Hidden Markov Models
- Kalman filter
- Markov decision processes, Reinforcement learning
- Gaussian Mixture Model (GMM), variational methods, mean field
- Variational auto-encoders
- Markov Chain Monte Carlo, Hamiltonian Monte Carlo
- Geometric Machine Learning: symmetries, groups, graph neural networks
- Attention / transformers
- Diffusion models, normalizing flow
- Nonlinear association measures
- Optimal transport
- Graph partitioning and network analysis
- Ethics of ML
- Q&A