initial setup

JuliaStats · Jan 27, 2013 · 240d9a4 · 240d9a4
1 parent 25a8254
commit 240d9a4
Show file tree

Hide file tree

Showing 3 changed files with 201 additions and 0 deletions.
diff --git a/LICENSE.md b/LICENSE.md
@@ -0,0 +1,7 @@
+Copyright (c) 2013 Dahua Lin
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
diff --git a/README.md b/README.md
@@ -0,0 +1,27 @@
+# Distances
+
+Evaluation of distances between vectors.
+
+## A List of supported distances
+
+* Minkowski distance
+* Euclidean distance
+* Squared euclidean distance
+* Cityblock distance 
+* Chebyshev distance
+* Hamming distance
+* Cosine distance
+* Spearman distance
+* Kullback-Leibler divergence
+* Jensen-Shannon divergence
+
+## Features
+
+* Many of the distances above accepts a weight vector as an optional argument to calculate weighted distances.
+* The module supports computation of distances in different ways:
+	- compute a distance between two vectors
+	- compute distances between a vector and an array comprised of multiple vectors
+	- compute distances between corresponding vectors in two arrays along a specific dimension
+	- compute distances between columns in two matrices in a pairwise manner
+* Specialized functions are used to compute pairwise (Squared) Euclidean distances in a much faster way.	
+
diff --git a/src/Distances.jl b/src/Distances.jl
@@ -0,0 +1,167 @@
+module Distances
+
+export 
+	GeneralizedDistance, 
+	Distance,
+	cdist,
+	pdist,
+	EuclideanDistance,
+	SquaredEuclideanDistance,
+	eucdist,
+	sqeucdist
+
+##########################################################################
+#
+#  The type hierarchy for distances
+#
+##########################################################################
+
+abstract GeneralizedDistance
+abstract Distance <: GeneralizedDistance
+
+type EuclideanDistance <: Distance 
+end
+
+type SquaredEuclideanDistance <: Distance 
+end
+
+##########################################################################
+#
+#  General cdist implementation
+#
+#  Remarks
+#  --------
+#  cdist(dist, a::Vector, b::Vector) should be implemented respectively
+#  for each distance.	
+#
+#  This general implementation extends them to matrices. 
+#
+##########################################################################
+
+function cdist(dist::GeneralizedDistance, a::AbstractVector, b::AbstractMatrix)
+	m, n = size(b)
+
+	# calculate the first distance (to determine value type)
+
+	d1 = cdist(dist, a, b[:,1])
+	r = Array(typeof(d1), n)
+	r[1] = d1
+
+	# calculate the rest
+
+	for i = 2 : n
+		r[i] = cdist(dist, a, b[:,i])
+	end
+
+	return r
+end
+
+function cdist(dist::GeneralizedDistance, a::AbstractMatrix, b::AbstractVector)
+	m, n = size(a)
+
+	# calculate the first distance (to determine value type)
+
+	d1 = cdist(dist, a[:,1], b)
+	r = Array(typeof(d1), n)
+	r[1] = d1
+
+	# calculate the rest
+
+	for i = 2 : n
+		r[i] = cdist(dist, sub(a, 1:m, i), b)
+	end
+
+	return r
+end
+
+function cdist(dist::GeneralizedDistance, a::AbstractMatrix, b::AbstractMatrix)
+	@assert size(a, 2) == size(b, 2)
+	n = size(a, 2)
+
+	# calculate the first distance (to determine distance value type)
+
+	d1 = cdist(dist, a[:,1], b[:,1])
+	r = Array(typeof(d1), n)
+	r[1] = d1
+
+	# calculate the rest
+
+	for i = 2 : n
+		r[i] = cdist(dist, a[:,i], b[:,i])
+	end
+
+	return r
+end
+
+
+##########################################################################
+#
+#  General pdist implementation
+#
+##########################################################################
+
+function pdist(dist::GeneralizedDistance, a::Matrix, b::Matrix)
+	m = size(a, 2)
+	n = size(b, 1)
+
+	# calculate the first distance (to determine distance value type)
+
+	d1 = cdist(dist, a[:,1], b[:,1])
+	r = Array(typeof(d1), (m, n))
+	r[1, 1] = d1
+
+	# calculate the rest
+
+	for i = 2 : m
+		b1 = b[:,1]
+		r[i, 1] = cdist(dist, a[:,i], b1)
+	end
+
+	for j = 2 : n
+		bj = b[:,j]
+		r[i, j] = cdist(dist, a[:,i], bj)
+	end
+
+	return r
+end
+
+##########################################################################
+#
+#  Specific distances
+#
+##########################################################################
+
+function cdist{T<:Real}(dist::EuclideanDistance, a::AbstractVector{T}, b::AbstractVector{T})
+	s::T = 0
+	n = length(a)
+	for i = 1 : n  # devectorize to make it faster
+		v = a[i] - b[i]
+		s += v * v
+	end
+	return sqrt(s)
+end
+
+function cdist{T<:Real}(dist::SquaredEuclideanDistance, a::AbstractVector{T}, b::AbstractVector{T})
+	s::T = 0
+	n = length(a)
+	for i = 1 : n  # devectorize to make it faster
+		v = a[i] - b[i]
+		s += v * v
+	end
+	return s
+end	
+
+
+##########################################################################
+#
+#  Convenient functions
+#
+##########################################################################
+
+eucdist(a::AbstractArray, b::AbstractArray)   = cdist(EuclideanDistance(), a, b)	
+sqeucdist(a::AbstractArray, b::AbstractArray) = cdist(SquaredEuclideanDistance(), a, b)
+
+
+end # module end	
+
+