|
I recently had a problem of having to define a distance metric between binary objects, and I knew a few metrics beyond the classics Jaccard, Minkowski, and so on. Then I came across the paper A survey of binary similarity and distance measures. It was very useful. Not only did it list over 70 metrics of similarity/distance but it calculated the correlation between them, so once I tried a particular metric that did not work for my purposes, I did not have to try the other metrics highly correlated to that. So it was a very useful knowledge - I not only wanted to share that with the community but also ask of other papers that share this same characteristic: listing and comparing a large set of techniques developed for the same purpose, so that it helps selecting one of these techniques. Thanks J. |
|
This isn't a particular paper suggestion, but often dissertations will have comprehensive comparative analyses as part of their background or even focus. Specifically in my area of interest, Active Learning, there is a good survey of the whole field, though it is a bit out of date (the author turned it into a full book that was recently published). Here is a link: http://www.cs.cmu.edu/~bsettles/pub/settles.activelearning.pdf |
I found another one very similar, but lists and correlates distances between distributions/histograms. http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.154.8446&rep=rep1&type=pdf