https://github.com/tmcclintock/covariance_breakdown

A small package to breakdown a covariance matrix into constituent parts useful for building emulators.
https://github.com/tmcclintock/covariance_breakdown

Last synced: 26 days ago
JSON representation

A small package to breakdown a covariance matrix into constituent parts useful for building emulators.

• Host: GitHub
• URL: https://github.com/tmcclintock/covariance_breakdown
• Owner: tmcclintock
• License: mit
• Created: 2018-12-13T20:41:18.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2019-03-19T17:34:22.000Z (almost 6 years ago)
• Last Synced: 2024-10-27T11:08:36.818Z (2 months ago)
• Language: Jupyter Notebook
• Homepage:
• Size: 375 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# covariance_breakdown

Constructing emulators of covariance matrices requires first breaking the matrix down into its constituent parts using a decomposition, and then rearranging diagonal and lower triangular data into a more useful form. This is a small package to do that.

Specifically, this packages takes in a covariance matrix, tests that it _can_ be decomposed, performs generalised Cholesky decomposition, and saves:

- the lower triangular matrix `L`

- the diagonal of the GCD `D` (which is in L)

- the flattened independent elements of L that aren't in D, `Lprime`

Additionally, the `breakdown` tool can reconstruct a covariance (and all the other parts) from `D` and `Lprime`.

## Usage

```python

#Given a covariance matrix C

from covariance_breakdown import *

C_breakdown = breakdown(C)

C_attr = C_breakdown.C #saves C as an attribute

D      = C_breakdown.D

L      = C_breakdown.L

Lprime = C_breakdown.Lprime

C_breakdown2 = breakdown.from_D_Lprime(D, Lprime)

#Has all the same attributes as C_breakdown

```