Modifications made on HammingANN class

[ghost]

Query bits were given same permutation as that of the training data bits and searching was done for each permutation of bits to find the approximate nearest neighbor.

[jakevdp]

I can't say I fully understand the change, and part of that is due to the fact that I haven't thought about this algorithm in a year and a half. Can you give more detail on what the problem was, and what you did to fix it?

[jakevdp]

THanks – that makes sense.

Would it be possible to add a unit test to explicitly check that this fix is working?

[ghost]

Yes, I will add a unit test to check the fix works.

[jakevdp]

Thanks! I'll look for your new commit.

[ghost]

Regarding unit test, I am not able to derive a relation between epsilon and accuracy. When you get time, can you share your thoughts on how it can be done ?