This repository is part of a lecture series given at Uppsala University, November 2020.
There are two directories, latex containing the presentation slides, and notebooks containing several jupyter notebooks.
In total there are 4+1 lectures, split into one half theory and one half coding each.
The goal of this series is to give participants hands on experience on how to utilize neural networks to tackle problems in string theory. We will reproduce results of a couple of recent papers and I'll also present some ongoing project(s) I'm currently working on.
The interested reader will find exercises and some open questions after each coding session. Hopefully these will inspire some to further dive in and investigate this rather new subfield of string theory.
Thanks to all the participants. If you have further questions to any of the notebooks or slides, don't hesitate to contact me. The last lecture will be given as part of the Uppsala journal club and will not include a coding session.
Monday 23.11 - 13.15:
Session 1 - Neural Networks
Application: Learning stability
Tuesday 24.11 - 13.15:
Session 2 - Regularization and CNNs
Application: Learning Hodge numbers
Wednesday 25.11 - 15.15:
Session 3 - Hyperparameter optimization
Application: Learning CY metrics
Thursday 26.11 - 13.15:
Session 4 - (Variational) Autoencoder
Application: Clustering of standard like models
Friday 04.12 - 13.15 (Journal club):
Session 5 - Reinforcement learning - Exploring standard like models
gymCICY tutorial
There are several nice reviews for machine learning from a (theoretical) physics perspective:
- Fabian Ruehle - Data science applications to string theory
- Jared Kaplan - Notes on Contemporary Machine Learning for Physicists
- Pankaj Mehta et al. - A high-bias, low-variance introduction to Machine Learning for physicists
Proper machine learning books:
- Christoper M. Bishop - Pattern Recognition and Machine Learning
- Ian Goodfellow and Yoshua Bengio and Aaron Courville - Deep Learning
There is an ongoing seminar series 'Physics meets ML' which can be found here. Before Christmas there will be the fourth (?) edition of string data. Magdalena is going to talk about our explore and exploit paper.