Data

Tools

Indexes

Applications

Experiments

Open Source Science

computationawarekalman.jl

Computation-Aware Kalman Filtering and RTS Smoothing

https://github.com/marvinpfoertner/computationawarekalman.jl

Science Score: 67.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ✓
    CITATION.cff file
    Found CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ✓
    DOI references
    Found 1 DOI reference(s) in README
  • ✓
    Academic publication links
    Links to: arxiv.org
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (6.6%) to scientific vocabulary

Keywords

kalman-filtering probabilistic-numerics rts-smoother
Last synced: 7 months ago · JSON representation ·

Repository

Computation-Aware Kalman Filtering and RTS Smoothing

Basic Info
  • Host: GitHub
  • Owner: marvinpfoertner
  • License: mit
  • Language: Julia
  • Default Branch: main
  • Homepage:
  • Size: 70.3 KB
Statistics
  • Stars: 9
  • Watchers: 2
  • Forks: 1
  • Open Issues: 0
  • Releases: 0
Topics
kalman-filtering probabilistic-numerics rts-smoother
Created over 2 years ago · Last pushed about 1 year ago
Metadata Files
Readme License Citation

README.md

ComputationAwareKalman.jl

ComputationAwareKalman.jl implements the computation-aware Kalman filter (CAKF) and the computation-aware RTS smoother (CAKS), novel approximate, probabilistic numerical versions of the Kalman filter and RTS smoother that are

  1. matrix-free and iterative, and can fully leverage modern parallel hardware (i.e. GPUs);
  2. more efficient than their standard versions, with quadratic time (worst-case) and linear memory complexities; and
  3. computation-aware, i.e. they come with theoretical guarantees for their uncertainty estimates which capture the inevitable approximation error.

In our paper we have demonstrated the scalability of the approach by applying it to a state-space model with $\approx 230\mathrm{k}$ dimensions in the context of spatiotemporal GP regression of climate/weather data with about $4$ million data points. The code for the experiments from the paper can be found in ComputationAwareKalmanExperiments.jl.

Citation

If you use this library, please cite our paper

bibtex @misc{Pfoertner2024CAKF, author = {Pf\"ortner, Marvin and Wenger, Jonathan and Cockayne, Jon and Hennig, Philipp}, title = {Computation-Aware {K}alman Filtering and Smoothing}, year = {2024}, publisher = {arXiv}, doi = {10.48550/arxiv.2405.08971}, url = {https://arxiv.org/abs/2405.08971} }

Owner

  • Name: Marvin Pförtner
  • Login: marvinpfoertner
  • Kind: user
  • Location: Tübingen, Germany
  • Company: University of Tübingen

PhD student in Machine Learning at the University of Tübingen

Citation (CITATION.cff) 

cff-version: 1.2.0
message: If you use this library, please cite it as below.
title: ComputationAwareKalman.jl
authors:
- name: Marvin Pförtner
license: MIT
url: "https://github.com/marvinpfoertner/ComputationAwareKalman.jl"
preferred-citation:
  type: generic
  title: "Computation-Aware Kalman Filtering and Smoothing"
  authors:
  - family-names: Pförtner
    given-names: Marvin
    orcid: "https://orcid.org/0000-0002-9005-2984"
  - family-names: Wenger
    given-names: Jonathan
    orcid: "https://orcid.org/0000-0003-2261-1331"
  - family-names: Cockayne
    given-names: Jon
    orcid: "https://orcid.org/0000-0002-3287-199X"
  - family-names: Hennig
    given-names: Philipp
    orcid: "https://orcid.org/0000-0001-7293-6092"
  year: 2024
  url: "https://arxiv.org/abs/2405.08971"
  identifiers:
    - type: other
      value: "arXiv:2405.08971"
      description: The arXiv preprint of the paper

GitHub Events

Total
  • Watch event: 8
  • Delete event: 3
  • Push event: 8
  • Fork event: 1
  • Create event: 3
Last Year
  • Watch event: 8
  • Delete event: 3
  • Push event: 8
  • Fork event: 1
  • Create event: 3