Tutorial on kernel ridge regression for the block course "Big data and artificial intelligence in materials science" as part of the Max Planck Graduate Center for Quantum Materials.
Tutorial on kernel ridge regression for materials science, created for the block course "Big data and artificial intelligence in materials science" as part of the [Max Planck Graduate Center for Quantum Materials](https://www.quantummaterials.mpg.de).
Topics covered:
- Kernel ridge regression
- Fundamentals
- HP Optimisation
- Hyper-Parameter Optimisation
- Toy implementation
- Representations
- Application to the NOMAD2018 challenge dataset
- Strategy
- Application to the [NOMAD2018 challenge dataset](https://www.kaggle.com/c/nomad2018-predict-transparent-conductors)
- Dataset exploration
- HP Optimisation
- Discussion of results
## Technicalities
In addition to the prerequisites in `setup.py`, this tutorial requires [`qmmlpack`](https://gitlab.com/qmml/qmmlpack/-/tree/development) on the DEVELOPMENT branch.
In addition to the prerequisites in `setup.py`, this tutorial requires [`qmmlpack`](https://gitlab.com/qmml/qmmlpack/-/tree/development) on the DEVELOPMENT branch, as well as [`cmlkit`](https://github.com/sirmarcel/cmlkit) and [`cscribe`](https://github.com/sirmarcel/cscribe).
It also requires the environment variable `CML_PLUGINS=cscribe` to be set.
Installing `qmmlpack` by hand, then `pip install cmlkit cscribe` should suffice.
## Todos/Future plans
- Once `cmlkit` supports caching, expand the result analysis section
