[FLINK-1933] Add distance measure interface and basic implementation to machine learning library

• Type: New Feature
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Component/s: Library / Machine Learning
• Labels: ML
• Assignee: Chiwan Park
• Reporter: Chiwan Park
• Created: 23/Apr/15 04:33
• Updated: 23/Jan/17 20:32
• Resolved: 07/May/15 14:17

Description

Add distance measure interface to calculate distance between two vectors and some implementations of the interface. In FLINK-1745, till.rohrmann suggests a interface following:

trait DistanceMeasure {
  def distance(a: Vector, b: Vector): Double
}

I think that following list of implementation is sufficient to provide first to ML library users.

• Manhattan distance [1]
• Cosine distance [2]
• Euclidean distance (and Squared) [3]
• Tanimoto distance [4]
• Minkowski distance [5]
• Chebyshev distance [6]

[1]: http://en.wikipedia.org/wiki/Taxicab_geometry
[2]: http://en.wikipedia.org/wiki/Cosine_similarity
[3]: http://en.wikipedia.org/wiki/Euclidean_distance
[4]: http://en.wikipedia.org/wiki/Jaccard_index#Tanimoto_coefficient_.28extended_Jaccard_coefficient.29
[5]: http://en.wikipedia.org/wiki/Minkowski_distance
[6]: http://en.wikipedia.org/wiki/Chebyshev_distance

Issue Links

• is required by: FLINK-1745 Add exact k-nearest-neighbours algorithm to machine learning library