Open jianshu93 opened 1 year ago
Hi Jianshu,
thank you for your questions regarding OMH.
Cheers
Hello Guillaume,
Thanks for the quick response during holiday! For the second question, my question then would be the distance/similarity estimated by OMH (the value calculated by omh_compute between 2 sequences, there must be a value telling whether high-probability poorly aligning sequence pairs and well aligning sequence pairs can be differentiated, like Figure 4 in the paper, correct me If I am understanding the paper in a wrong way) metric? Thanks for pointing out the indik's definition. The reason I am asking is also related to nearest neighbor search, but not based on LSH, but based on tree-like or graph like space partitioning algorithms, which requires that the distance used must be metric. I want to feed the output from omh_compute (a distance) to graph or tree like algorithms like this one here, which requires the distance used must be metric (or at least close).
Thanks,
Jianshu
Question on to what edit dissimilarity between 2 sequences, order Min Hash is still accurate. I can see that for 0.45 edit dissimilarity, it is still accurate and can differentiate those 2 sequences. Another question is, if 2 sequences are not the same length, say one is only 1/5 of another, then will order min hash can still accurately accproxiamte so called, semi-global alignment identity?
Thanks,
Jianshu
Hello Team,
several questions related to omh and edit distance:
Thanks,
Jianshu