I think I was the source of this expectation. And I also think I was wrong.
I just did some experiments myself in R and random cut vectors seem to work about as well as non-random ones for positive orthant vectors. For oddly distributed vectors, it still might be good to use difference vectors as a basis for LSH, but I am much less convinced than before. On Sun, May 15, 2011 at 8:54 PM, Lance Norskog <[email protected]> wrote: > Test data: 1000 random vectors as samples. All values 0->1, linear > distribution. > This test data gives no negative cosine distances, and so all bits are > 0. This is expected (from previous mail threads). >
