Can Locality Sensitive Hashing be applied on dynamic-dimensional data points?

Question

For example assume that we have some vectors with differnt length and what we want to do is measuring the similarity between each two pair of these vectors. What we have to consider is that these vectors' dimensions are time-varying. Can we do this?

Answer 1

After having some research about the topic, I found some note to share with you. The disadvantages of LSH is as follows.

• Because LSH needs to use a large number of hash tables, it consumes a large amount of memory.
• It cannot also simply adapt to data insertion and deletion.
• Moreover, not all similarity metrics have a suitable hash function.
• Afterwards, LSH highly depends on the length of vectors. There might be some situation in which two vectors are similar to each other even a little bit but LSH cannot find any Minhash in common, and therefore these vector are not considered as similar.

I hope be helpful.