Matrix M
is the signatures matrix, which is produced via Minhashing of the actual data, has documents as columns and words as rows. So a column represe
I think I figured it out, posting for future readers.
I am going to use one dictionary, since the slides mentioned that it's OK to use the same hash function for every stripe (dictionaries do that).
Every bucket will be a key for our dictionary.
On insertion, a document (i.e. a column which belongs in a stripe) will be passed by a hash function (which we will create) and the result should be a key. That way our dictionary will be populated.