Notes on “WaveLab and Reproducible Research”

October 2, 2008

WaveLab and Reproducible Research: Jonathan B. Buckheit and David. L. Donoho.

The Gist: Science itself rests on the reproducibility of published experiments. Computational science results must be reproducible. The authors allow the easy reproduction of their wavelet experiments with the release of a software package called WaveLab. 

The ideal journal publication simplifies reproduction. Computational science journals should improve themselves and require reproducibility of their papers.

The good: The authors lead by example. First, they explain the problem to be solved in self-deprecating terms. Then, they demonstrate how they solve the problem within their domain. They also show off some of the power of their (obscure?) area.

Aside: I’m glad for this week’s batch of papers. They’re good. I was starting to become cynical about the state of science in the software world.

Advertisements

One Response to “Notes on “WaveLab and Reproducible Research””

  1. Ian Says:

    Sounds like a good read.

    Wavelets are actually the basis for the JPEG2000 standard – the original JPEG (published in 1994) was based on 2-dimensional DCTs. Wavelets are suppose to have better characteristics for compressing than the DCT. They were proposed for the original JPEG, but were not sufficiently developed in the early 80’s to be adopted by the standards committee.


Comments are closed.

%d bloggers like this: