# Collection of Resources: Deep Learning for Physiscs

Dear enthusiasts, Phyisics Informed Deep Learning is a rather new but very interesting field. There is actually already a quite exhaustive collection of papers/datasets/projects out there which you can find on this [Physics-Based Deep Learning GitHub Repository](https://github.com/thunil/Physics-Based-Deep-Learning).
The following things are some addations to this.

---
## Talks and Blogs
* [Miles Cramner: Interpretable Deep Learning for New Phyisics Discovery](https://youtu.be/HKJB0Bjo6tQ)
* [Yannic Kilcher about Cramners Paper](https://youtu.be/LMb5tvW-UoQ)
* [Cramner: Lagrangian Neural Networks](https://www.youtube.com/watch?v=27ravidF96g&t=3019s)
---

## Papers

### Overview Papers
* [Integrating Physics-Based Modeling With Machine
Learning: A Survey](https://beiyulincs.github.io/teach/fall_2020/papers/xiaowei.pdf)
* [Physics-informed Machine Learning](https://www.brown.edu/research/projects/crunch/sites/brown.edu.research.projects.crunch/files/uploads/Nature-REviews_GK.pdf)
* [Theory-Guided Data Science: A New Paradigm for Scientific Discovery from Data](https://arxiv.org/pdf/1612.08544.pdf)
* 2022: [Scientific Machine Learning Through Physics–Informed Neural Networks: Where we are and What’s Next](https://link.springer.com/article/10.1007/s10915-022-01939-z)

### Methods
* [Neural Partial Differential Equations for Chaotic Systems](https://iopscience.iop.org/article/10.1088/1367-2630/abeb90/pdf)
* [Physics-informed neural networks: A deep learning
framework for solving forward and inverse problems involving
nonlinear partial differential equations](https://faculty.sites.iastate.edu/hliu/files/inline-files/PINN_RPK_2019_1.pdf)
* [Lagrangian Neural Networks](https://arxiv.org/pdf/2003.04630.pdf)
* [Fourier Neural Operator for Parametric Partial Differential Equations](https://arxiv.org/abs/2010.08895)
* [Differentiable Physics-informed Graph Networks](https://arxiv.org/pdf/1902.02950.pdf)
* [Physics-informed graph neural Galerkin networks: A unified framework for solving PDE-governed forward and inverse problems](https://www.sciencedirect.com/science/article/abs/pii/S0045782521007076)
* [PhyGNNet: Solving spatiotemporal PDEs with Physics-informed Graph Neural Network](https://arxiv.org/abs/2208.04319)
*
---

## GitHub Repositories
* [Physic's Informede Neural Networks Introduction](https://maziarraissi.github.io/PINNs/)
* [Physics-Based Deep Learning](https://github.com/thunil/Physics-Based-Deep-Learning) - collection of materials (papers, data)
* [Lagrangian Neural Networs](https://github.com/MilesCranmer/lagrangian_nns)
---

## Tutorials
* [NIPS 20201: ML for Physics and Physisc for ML](https://nips.cc/virtual/2021/tutorial/21896) - you need to register but it's free
---

## Data and Libraries
* [DeepXDE - deep learning for solving differential equations](https://deepxde.readthedocs.io/en/latest/)
* [Physics Informed Neural Networks](https://github.com/maziarraissi/PINNs) - some data and code
---

## Extra Section: Physic-driven ML and earth sciences
This list is a work in progress (some papers might not fit well here and a lot of papers are missing).

### Overview
* [Deep learning and process understanding for data-driven Earth system science](https://www.nature.com/articles/s41586-019-0912-1)
* [Physics-informed machine learning: case studies for weather and climate modelling](https://royalsocietypublishing.org/doi/10.1098/rsta.2020.0093) - here is also a [YouTube presentation](https://www.youtube.com/watch?v=e2CCUscL_ok&ab_channel=USCLIVAR) by them
* [Climbing down Charney’s ladder: machine learning and the post-Dennard era of computational climate science](https://royalsocietypublishing.org/doi/full/10.1098/rsta.2020.0085#d1e1007) - that's the history part

### Causality
* Very informal introduction to Causal ML: [Why we need causality in ML](https://towardsdatascience.com/introduction-to-causality-in-machine-learning-4cee9467f06f)
* [Inferring causation from time-series in Earth system sciences](https://www.nature.com/articles/s41467-020-15195-y) - a bit more physics and climate model related example
* [Causal inference for climate change events from satellite image time series using computer vision and deep learning](https://arxiv.org/abs/1910.11492) - not climate model related example
* Neurips 2019 Competition: [The causality for climate competition](https://proceedings.mlr.press/v123/runge20a.html)

### Downscaling
The higher resolution, the better. No matter if we speak about climate models, satellite data, etc.
* [Configuration and intercomparison of deep learning neural models for statistical downscaling](https://gmd.copernicus.org/articles/13/2109/2020/)
* ...

### Emulators
Emulate physics-based model. Can be any model, here often referring to climate models. Accelerates the physical model, could provide more fine-grained output and in climate models: more scenarios.
* (simple overview paper needs to be added)
* [Calibrate, emulate and sample](https://www.sciencedirect.com/science/article/pii/S0021999120304903) - more on the math side
* [ClimateBench](https://www.essoar.org/doi/10.1002/essoar.10509765.2) - emulates temp/precip of an ESM (earth system model); dataset included
* [ClimArt](https://arxiv.org/abs/2111.14671) - emulates radiative fluxes of an ESM; dataset included
* *(There are incredibly many emulators out there, not sure yet which ones a good first examples)*

### Other