https://github.com/alexbuccheri/isdf_prototypes

Prototype code for Interoperable Separable Density Fitting (ISDF) of 2-e Integrals
https://github.com/alexbuccheri/isdf_prototypes

Last synced: about 2 months ago
JSON representation

Prototype code for Interoperable Separable Density Fitting (ISDF) of 2-e Integrals

• Host: GitHub
• URL: https://github.com/alexbuccheri/isdf_prototypes
• Owner: AlexBuccheri
• Created: 2024-04-06T18:19:40.000Z (10 months ago)
• Default Branch: main
• Last Pushed: 2024-11-24T18:56:38.000Z (2 months ago)
• Last Synced: 2024-11-24T19:37:42.358Z (2 months ago)
• Language: Jupyter Notebook
• Size: 412 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# ISDF Prototypes

This project is for prototyping algorithms associated with Inter Separable Density Fitting (ISDF), with the aim of

speeding up the evaluation of two-centre integrals $(ij|kl)$

## Algorithms

* [Weighted k-means clustering for finding interpolation points](src/isdf_prototyping/interpolation_points.py)

## Papers of Relevance

* K-means Clustering:

  * J. Chem. Theory Comput. 2018, 14, 1311−1320

  * Complex-Valued K-means Clustering for Interpolative Separable Density Fitting to Large-Scale Hybrid Functional 

    Ab Initio Molecular Dynamics with Plane-Wave Basis Sets

## Websites of Relevance

* Basics

  * [Basics of 2-electron integrals](http://vergil.chemistry.gatech.edu/notes/permsymm/permsymm.pdf)

* K-means Clustering:

  * [Google tech talk](https://www.youtube.com/watch?v=NDAVDRFMh_0) on k-means algorithms

## Implementation Notes

### Weighted K-means and Finding Interpolation Points

**Issue with My Implementation**

SKLearn implementation does better than my implementation, even without using the k-means++ algorithm for better 

initial guess at the centroids. This suggests that there's an issue in my implementation

 

_**TODO:**_ Debug

SK Learn either uses greedy k-means++ (which I have not yet implemented) or if using random centroids, it runs 10 times

and selects the points that are most optimised.

_**TODO:**_ Either implement greedy k-means++ or have the random seeding algorithm run ~ 10 times and choose the best set of

centroids.

**Implement QD Decomposition as a Reference**

Should also implement QR decomposition for comparison