loess 0.0.4 → 1.0.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 57c70d6636f7fd381261a7af28fdb5a0765e8bc0
-  data.tar.gz: c76458cf22700ecfbadad9582c9ae16a2f19d3b7
+  metadata.gz: 505e1cc10dcad7f56fd80a8743e51bf27c188a19
+  data.tar.gz: 2162e9cbf0faef7811692fa24e7523b4cc79fce7
 SHA512:
-  metadata.gz: ea7dc5e99f00a2db6ceafc44f624e303b0c11a29f1beedb58e960fc66c56f766d52c5b2bb188387e08d376fd0260646ce82d711a178de989c866f165f69ef370
-  data.tar.gz: 06fc4e3dab6612eeff5085d30ba2e850a0df7a62486cb1e4298d76e049e04df1c119863e8f648f9fa8a33d03f81d711faf64be4a0df6da51455c61721dc8823a
+  metadata.gz: c90e2002e56320470697d39a661dd8142bd347485c7404a1434adc087df3f2813aa8bfbb4416852c6dd5c94e507ee7f06c733510cb37969b598ff30293a40c98
+  data.tar.gz: e8b439a545b23f08e17c5e3bdf31d7255c328b4752444c0b4ec2d8afd8e4977e08ff06af854abeb4f3a0af918a52cd60ce4f52fa0ee9f27cbc6317fb9fba0401

data/README.md

@@ -1,6 +1,6 @@
 # Loess
 
-Simple Loess / Lowess interpolator built in Ruby, based on Apache's [`LoessInterpolator`][1].
+Simple Loess / Lowess interpolator built in Ruby, using Apache's [`LoessInterpolator`][1] through Rjb
 
 ## Installation
 
@@ -16,31 +16,41 @@ Or install it yourself as:
 
     $ gem install loess
 
+## Depencies
+
+This gem depends on Rjb, which in turn depends on Java. [Rjb documentation][3] has more info.
+
 ## Usage
 
 ```Ruby
 
 data = 1000.times.map { |i| [i, rand(10_000)] }
 
-regression = Loess::Calculator.calculate(data)
-regression = Loess::Calculator.new(data).calculate
+regression = Loess::Interpolator.interpolate(data)
+regression = Loess::Interpolator.new(data).interpolate
 
 # Change your settings
 
-calculator = Loess::Calculator.new(data)
-calculator.bandwidth = 0.2
-calculator.robustness_factor = 10 # Go crazy
-calculator.calculate
+interpolator = Loess::Interpolator.new(data)
+interpolator.bandwidth = 0.2
+interpolator.robustness_factor = 10 # Go crazy
+interpolator.interpolate
 
 # Pass in settings through initialize
-calculator = Loess::Calculator.new(data, bandwidth: 0.2, accuracy: 1e-10)
-calculator.calculate
+interpolator = Loess::Interpolator.new(data, bandwidth: 0.2, accuracy: 1e-10)
+interpolator.interpolate
 
+# Get a Spline Interpolator object
+# Returns an instance of PolynomialSplineFunction, smoothed with the given data
+# More documentation at [Apache Commons Maths][2]
+interpolator.spline_interpolator
 
 ```
 
 ## Contributing
 
+At this moment, there is little to contribute, but still:  
+
 1. Fork it ( https://github.com/swanandp/loess.rb/fork )
 2. Create your feature branch (`git checkout -b my-new-feature`)
 3. Commit your changes (`git commit -am 'Add some feature'`)
@@ -48,4 +58,6 @@ calculator.calculate
 5. Create a new Pull Request
 
 
-[1]: http://commons.apache.org/proper/commons-math/jacoco/org.apache.commons.math3.analysis.interpolation/LoessInterpolator.java.html
+[1]: http://commons.apache.org/proper/commons-math/jacoco/org.apache.commons.math3.analysis.interpolation/LoessInterpolator.java.html
+[2]: http://commons.apache.org/proper/commons-math/jacoco/org.apache.commons.math3.analysis.polynomials/PolynomialSplineFunction.java.html
+[3]: https://github.com/arton/rjb

data/lib/loess.rb

@@ -1,10 +1,9 @@
 # -*- encoding : utf-8 -*-
+require 'rjb'
+Rjb::add_jar(File.join(File.expand_path('../../', __FILE__), 'vendor', 'apache', 'commons-math3-3.3.jar').to_s)
 
-# A Ruby Port of Apache's LoessInterpolator
-# http://commons.apache.org/proper/commons-math/jacoco/org.apache.commons.math3.analysis.interpolation/LoessInterpolator.java.html
-#
 module Loess
-  class Calculator
+  class Interpolator
     attr_accessor :data, :bandwidth, :robustness_factor, :accuracy
     # Sane Defaults as per Apache
     DEFAULT_ACCURACY          = 1e-12
@@ -16,125 +15,31 @@ module Loess
     # For options, refer to defaults above
     def initialize(data = [], options = {})
       @xval, @yval           = split_up(data)
-      self.accuracy          = options[:accuracy] || DEFAULT_ACCURACY
-      self.bandwidth         = options[:bandwidth] || DEFAULT_BANDWIDTH
-      self.robustness_factor = options[:robustness_factor] || DEFAULT_ROBUSTNESS_FACTOR
+      self.accuracy          = Float(options[:accuracy] || DEFAULT_ACCURACY)
+      self.bandwidth         = Float(options[:bandwidth] || DEFAULT_BANDWIDTH)
+      self.robustness_factor = Integer(options[:robustness_factor] || DEFAULT_ROBUSTNESS_FACTOR)
+      @klass                 = Rjb::import('org.apache.commons.math3.analysis.interpolation.LoessInterpolator')
+      @interpolator          = @klass.new(bandwidth, robustness_factor, accuracy)
     end
 
     # Accepts array of [x, y] pairs
     # e.g. [ [1, 2], [3, 4], [5, 6], [0, 42] ]
-    def calculate(data = nil)
+    def interpolate(data = nil)
       @xval, @yval = split_up(data) if data
-      smooth(@xval, @yval)
+      @interpolator.smooth(@xval, @yval)
     end
 
-    def smooth(xval, yval, weights = [])
-      xlength = xval.length
-      return unless xlength > 0 && xlength == yval.length
-      return yval if xlength == 1 || xlength == 2
-      bandwidth_in_points = (bandwidth * xlength).to_i
-      fail "bandwidth is way too small" if bandwidth_in_points < 2
-      weights = Array(weights).length > 0 ? weights : [1.0] * xlength
-
-      result             = []
-      residuals          = []
-      robustness_weights = [1] * xlength
-
-      robustness_factor.times do |factor|
-        xval.each_with_index do |x, i|
-          bandwidth_interval = [0, bandwidth_in_points - 1]
-          update_bandwidth_interval(xval, weights, i, bandwidth_interval) if i > 0
-          ileft, iright = bandwidth_interval
-          edge          = if xval[i] - xval[ileft] > xval[iright] - xval[i]
-                            ileft
-                          else
-                            iright
-                          end
-
-          sum_weights   = 0
-          sum_x         = 0
-          sum_x_squared = 0
-          sum_y         = 0
-          sum_xy        = 0
-          denom         = (1.0 / (xval[edge] - x)).abs
-
-          xval[ileft..iright].each_with_index do |xk, k|
-            next unless xk
-            yk            = yval[k]
-            dist          = (k < i) ? x - xk : xk - x
-            w             = tricube(dist * denom) * robustness_weights[k] * weights[k]
-            xkw           = xk * w
-
-            # Intentionally avoiding multiple reduce calls here
-            # On large data-sets, this severly impacts performance
-            sum_weights   += w
-            sum_x         += xkw
-            sum_x_squared += xk * xkw
-            sum_y         += yk * w
-            sum_xy        += yk * xkw
-          end
-
-          mean_x         = sum_x / sum_weights
-          mean_y         = sum_y / sum_weights
-          mean_xy        = sum_xy / sum_weights
-          mean_x_squared = sum_x_squared / sum_weights
-
-          beta = if Math.sqrt((mean_x_squared - mean_x * mean_x).abs) < accuracy
-                   0
-                 else
-                   (mean_xy - mean_x * mean_y) / (mean_x_squared - mean_x * mean_x)
-                 end
-
-          alpha        = mean_y - beta * mean_x
-          result[i]    = beta * x + alpha
-          residuals[i] = (yval[i] - result[i]).abs
-        end
-
-        break if factor == robustness_factor - 1
-
-        sorted_residuals = residuals.sort
-        median_residual  = sorted_residuals[xlength / 2]
-
-        break if median_residual.abs < accuracy
-
-        xlength.times do |i|
-          arg = residuals[i] / (6 * median_residual)
-          if arg >= 1
-            robustness_weights[i] = 0
-          else
-            robustness_weights[i] = (1 - arg * arg) ** 2
-          end
-        end
-      end
-      result
-    end
-
-    # http://en.wikipedia.org/wiki/Local_regression#Weight_function
-    def tricube(x)
-      (1 - x.abs ** 3) ** 3
-    end
-
-    def update_bandwidth_interval(xval, weights, i, bandwith_interval)
-      left, right = bandwith_interval
-      next_right  = next_non_zero(weights, right)
-      if next_right < xval.length &&
-          xval[next_right] - xval[i] < xval[i] - xval[left]
-        next_left            = next_non_zero(weights, left)
-        bandwith_interval[0] = next_left
-        bandwith_interval[1] = next_right
-      end
-    end
-
-    def next_non_zero(collection, index)
-      collection.each_with_index.detect { |el, i|
-        i > index && !el.zero?
-      }[1]
+    # Accepts array of [x, y] pairs
+    # e.g. [ [1, 2], [3, 4], [5, 6], [0, 42] ]
+    def spline_interpolator(data = nil)
+      @xval, @yval = split_up(data) if data
+      @interpolator.interpolator(@xval, smooth(@xval, @yval))
     end
 
     # Accepts array of [x, y] pairs
     # e.g. [ [1, 2], [3, 4], [5, 6], [0, 42] ]
-    def self.calculate(data)
-      new(data).calculate
+    def self.interpolate(data, options = {})
+      new(data, options).interpolate
     end
 
     private

data/lib/loess/version.rb

@@ -1,3 +1,3 @@
 module Loess
-  VERSION = "0.0.4"
+  VERSION = "1.0.0"
 end

data/loess.gemspec

@@ -8,7 +8,7 @@ Gem::Specification.new do |spec|
   spec.version       = Loess::VERSION
   spec.authors       = ["Swanand Pagnis"]
   spec.email         = ["swanand.pagnis@gmail.com"]
-  spec.summary       = %q{A Simple LOESS / LOWESS calculator built in Ruby}
+  spec.summary       = %q{A Simple LOESS / LOWESS calculator built in Ruby, using Apache Commons Math through Rjb}
   spec.description   = %q{Perfect if you want to plot a line graph or scatter plot and a loess regression}
   spec.homepage      = "https://github.com/swanandp/loess.rb"
   spec.license       = "MIT"
@@ -20,4 +20,6 @@ Gem::Specification.new do |spec|
 
   spec.add_development_dependency "bundler", "~> 1.6"
   spec.add_development_dependency "rake"
+
+  spec.add_runtime_dependency "rjb"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: loess
 version: !ruby/object:Gem::Version
-  version: 0.0.4
+  version: 1.0.0
 platform: ruby
 authors:
 - Swanand Pagnis
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2014-08-05 00:00:00.000000000 Z
+date: 2014-08-08 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bundler
@@ -38,6 +38,20 @@ dependencies:
     - - ">="
       - !ruby/object:Gem::Version
         version: '0'
+- !ruby/object:Gem::Dependency
+  name: rjb
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: '0'
+  type: :runtime
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: '0'
 description: Perfect if you want to plot a line graph or scatter plot and a loess
   regression
 email:
@@ -55,6 +69,7 @@ files:
 - lib/loess.rb
 - lib/loess/version.rb
 - loess.gemspec
+- vendor/apache/commons-math3-3.3.jar
 homepage: https://github.com/swanandp/loess.rb
 licenses:
 - MIT
@@ -78,5 +93,6 @@ rubyforge_project:
 rubygems_version: 2.2.2
 signing_key: 
 specification_version: 4
-summary: A Simple LOESS / LOWESS calculator built in Ruby
+summary: A Simple LOESS / LOWESS calculator built in Ruby, using Apache Commons Math
+  through Rjb
 test_files: []