What is the number of Hash functions needed in Bundle Min Hashing for Logo Recognition?

Asked Oct 28 '14 at 06:39

Active Aug 03 '18 at 07:45

Viewed 369 times

In reference to the paper Bundle Min Hashing for Logo Recognition:

Suppose we have bundles {2,5,18,444,678} and {2,5,79,368,841} and the vocabulary size is 1M. If we have just 1 sketch per bundle then do we need just 1 hash function which hashes 1M integers deterministically to values from uniform distribution in [0,1]. The hash function must have fixed seed for each call.For 4-sketches we just need the same hash function with 4 seeds. Is the thought correct?

Or can we randomly pick a number from the set(bundle) as Min Hash word, since they represent random permutation of set?

Any reference for implementation of hash functions needed in the paper?

Can MurmurHash3 do the work?

asked Oct 28 '14 at 06:39

Ankit Nayan

1

BTW I have implemented the paper with amazingly similar accuracy. If somebody needs this technology, I shall outsource with permissions. – Ankit Nayan Sep 19 '16 at 13:55
I need a good implementation of this . Can you outsource it? – MysticForce Aug 17 '17 at 11:28
shared the implementation at https://github.com/ankitnayan/bundle_min_hashing_source – Ankit Nayan Aug 22 '18 at 05:29

What is the number of Hash functions needed in Bundle Min Hashing for Logo Recognition?

0 Answers0