axiomzen / alike Goto Github PK

But should punish them 👊

cc @mck-

Saw your library on HN. Looks useful. Quick very minor suggestion: you say:

Where movies is an array of objects that have at least those 5 attributes. Returns the top
10 movies from the array. 

You should fill out this example with that sample data in movies so we could see it actually work. It wouldn't add much bloat to the README file and would lead to a much fuller understanding of the behavior of the library.

Thanks for making this available,
t

The idea is that all results matched would be deep copied, can it be done? Would be awesome =)

consider this example:

subject = {proximity: 1, updated_time: 0};
objects = [
  {proximity: 0.9, updated_time: 1413345215},
  {proximity: 0.7, updated_time: 1413345115}
];

if subject is not used in calculation of the stdv the distances of objects from the subject will be skewed by dimensions that have consistently larger values with respect to those of the subject.
according to here when calculating the distances in a sample, all members of the sample are used to calculate the stdv

They seems to be doing the same thing. Can we merge the commits, and keep only one? I think we have a limit on #of private repos at the moment ;)
https://github.com/axiomzen/SimilarityLib

Also, if it's ready to be used by others, maybe we should open source this?

cc: @mck- @baugarten

Hi,

Thank you for your work and interesting library! My question is:
Is it possoble to get index of matched neighbours instead of array of matched attributes?

Based on README example I would like to get index (id) of a movie(s) insetad of listed attributes.

Many thanks,
Tomek

I was trying to use the debug option but it doesn't exist. Looks like the NPM package is out of date and doesn't include the commit. Would be great if it was updated.

Great stuff in addition to the look-alike module. I was wondering would it be possible to return the accuracy/match for results?

thanks

@mck-
I've found it on many occasions very useful to have a "debug" option that if set to true would return distances along with the results. it helps with understanding why an object is ranked the way it's ranked when debugging. also in some cases user may need the actual distances as well as the ranking.

utils.distance({ a: 1, b: 30 }, { a: 1, b: 50 }, { stdv: { a: 0, b: 10 } }) 

returns null