Re:Once you go functional, you can never go back
by vidarh on Monday May 18, 2015 @10:39AM
(#9564682)
If only they did define these things precisely and rigorously. Often it's sloppier than a lot of pseudo-code I see.
E.g. I did my MSc on reducing error rates in OCR through various statistical methods. As part of that I needed to review a lot of papers applying various filters to input images. I went through at least a dozen papers on thresholding (filtering out noise by excluding all pixels under a certain brightness, for example)
Nearly all the papers had formulas. In this case it was fine. A simple-stupid (and not very good) thresholding function is simply (pseudo code): f(color) = intensity(color) > threshold ? color : 0. It's hard to obscure that by changing the notation - it'll be obvious in most cases what such a simple function does.
But nearly all of the results depended on specific values for specific variables in those formulas, and nearly all of them left out sufficient information about the values that worked best (or worked all) and/or did what I did above: define another function - intensity() in my case - that was left undefined, even though what method you'd use to determine intensity would have drastic impact on the results.
E.g. I did my MSc on reducing error rates in OCR through various statistical methods. As part of that I needed to review a lot of papers applying various filters to input images. I went through at least a dozen papers on thresholding (filtering out noise by excluding all pixels under a certain brightness, for example)
Nearly all the papers had formulas. In this case it was fine. A simple-stupid (and not very good) thresholding function is simply (pseudo code): f(color) = intensity(color) > threshold ? color : 0. It's hard to obscure that by changing the notation - it'll be obvious in most cases what such a simple function does.
But nearly all of the results depended on specific values for specific variables in those formulas, and nearly all of them left out sufficient information about the values that worked best (or worked all) and/or did what I did above: define another function - intensity() in my case - that was left undefined, even though what method you'd use to determine intensity would have drastic impact on the results.