loganb-kder 0.1.0

checksums.yaml

@@ -0,0 +1,7 @@
+---
+SHA1:
+  metadata.gz: cde53f25f8badc44aef6aaf29bcb0d255940fd73
+  data.tar.gz: 3568e1d009ba8bc7fac4a610a3d0882349818bbd
+SHA512:
+  metadata.gz: b44e25490f96342d66f9c70ba8e3386372380dc086e5beda9ad0f0e6784b095b9f50042ef28442a2e14f7b5427737527aa7336269c3945df5079e41f594b8137
+  data.tar.gz: 112de7099a77c778a60d25468abf1890c7990db6f086771aa80b7cd4206b3313c397f948d8f1f92526460e9695d998e07f304173df509a359da32352e3a272ad

data/lib/kder/bandwidth.rb

@@ -0,0 +1,13 @@
+require_relative '../util/enum_extensions'
+module Bandwidth
+  class << self
+    def silverman(input_arr, std_deviation = input_arr.standard_deviation) 
+      bandwidth = ((4 * std_deviation ** 5)/(3 * input_arr.size))**(1.0/5)
+    end
+    alias_method :estimate, :silverman 
+    alias_method :nrd0, :silverman
+
+    #def mean_integrated_squared_error(a,b,c,d)
+    #end
+  end
+end

data/lib/kder/kde.rb

@@ -0,0 +1,86 @@
+require 'set'
+
+class Kder
+  require_relative 'bandwidth'
+  require_relative '../util/statistics'
+
+  Sigmas = 2.5 
+  MeshCount = 2e3
+  MinimumThresholdValue = 1e-2
+  MinimumStepSize = 1e-3
+  DifferenceThreshold = 1e-3
+  class << self
+    ## 
+    # :singleton-method: kde
+    # Accepts a single member array plus optional additional information
+    # Returns a two member array, [x_vals,y_vals] representing the kde
+    def kde(arr, bw = nil, opts = {sigmas: Sigmas, sampling_density: MeshCount, threshold: MinimumThresholdValue, minimum_delta: DifferenceThreshold})
+      unless bw # is nil
+        bw = Bandwidth.silverman(arr)
+      end
+      bw = bw == 0 ? 0.1 : bw
+      # Initialization steps
+      range = bw*opts[:sigmas]
+      min = arr.min - range
+      max = arr.max + range
+      step_size = (max-min)/(opts[:sampling_density].to_f)
+      step_size = step_size < MinimumStepSize ? MinimumStepSize : step_size
+      arr.sort!
+      # initialize the range variables
+      ranges = (min..max).step(step_size).to_a
+      output = [[min,0]]
+      old_intensity = 0
+      # Step through the range
+      ranges[1..-1].map.with_index do |mid, i|
+        high_end = mid + range
+        lower_end = mid - range
+        selection_range = (lower_end..high_end)
+        included = arr.select {|a| selection_range.include?(a)}
+        intensity = included.map {|a| Kder::Statistics.custom_pdf(a-mid, bw) }.inject(:+) || 0
+        unless intensity < opts[:threshold] or (intensity - old_intensity).abs < opts[:minimum_delta]
+          output << [mid, intensity ] 
+          old_intensity = intensity
+        end
+      end
+      output << [max,0]
+      output.compact.transpose
+    end
+
+    #
+    # :singleton-method: kdevec
+    # Works just like kdevec, but arr is a list of 2-D points of sample value and magnitude. Is this valid math? ¯\_(ツ)_/¯
+    # 
+    def kdevec(arr, bw = nil, opts = {}.freeze)
+      opts = {sigmas: Sigmas, sampling_density: MeshCount, threshold: MinimumThresholdValue, minimum_delta: DifferenceThreshold}.merge(opts)
+      raise "bandwidth must be specified for now" unless bw
+
+      values = arr.sort_by { |v| v[0] }
+
+      # Initialization steps
+      range = bw*opts[:sigmas]
+      min = values.first[0] - range
+      max = values.last[0]  + range
+      step_size = (max-min)/(opts[:sampling_density].to_f)
+      step_size = step_size < MinimumStepSize ? MinimumStepSize : step_size
+
+      # initialize the range variables
+      ranges = (min..max).step(step_size).to_a
+      output = [[min,0]]
+      old_intensity = 0
+      # Step through the range
+      ranges[1..-1].map.with_index do |mid, i|
+        high_end  = mid + range
+        lower_end = mid - range
+        selection_range = (lower_end..high_end)
+        included = values.select {|v| selection_range.include?(v[0])}
+        intensity = included.map {|v| v[1] * Kder::Statistics.custom_pdf(v[0]-mid, bw) }.inject(:+) || 0
+        unless intensity < opts[:threshold] or (intensity - old_intensity).abs < opts[:minimum_delta]
+          output << [mid, intensity ] 
+          old_intensity = intensity
+        end
+      end
+      output << [max,0]
+      output.compact.transpose      
+    end
+  end
+end

data/lib/kder/version.rb

@@ -0,0 +1,3 @@
+class Kder
+  VERSION = "0.1.0"
+end

data/lib/kder.rb

@@ -0,0 +1 @@
+require_relative 'kder/kde'

data/lib/util/enum_extensions.rb

@@ -0,0 +1,19 @@
+module Enumerable
+  # Thanks StackOverflow users!
+  # http://stackoverflow.com/questions/7749568/how-can-i-do-standard-deviation-in-ruby
+  def sum
+    self.inject(:+)
+  end
+  def mean
+    self.sum/self.length.to_f
+  end
+  alias average mean
+  def sample_variance
+    m = self.mean
+    sum = self.inject(0){|accum, i| accum + (i-m)**2}
+    sum/(self.length - 1).to_f
+  end
+  def standard_deviation
+    Math.sqrt(self.sample_variance)
+  end
+end

data/lib/util/statistics.rb

@@ -0,0 +1,10 @@
+class Kder
+  module Statistics
+    SQ2PI = Math.sqrt(2 * Math::PI)
+    # Normal pdf with a custom sigma value, still centered around 0.
+    def self.custom_pdf(x, sigma = 1.0)
+      mean = 0 
+      (1.0/( sigma*SQ2PI))*Math::exp(-0.5*((x-mean)/sigma)**2)
+    end
+  end
+end

metadata

@@ -0,0 +1,53 @@
+--- !ruby/object:Gem::Specification
+name: loganb-kder
+version: !ruby/object:Gem::Version
+  version: 0.1.0
+platform: ruby
+authors:
+- Ryan Taylor
+- Logan Bowers
+autorequire: 
+bindir: bin
+cert_chain: []
+date: 2017-11-15 00:00:00.000000000 Z
+dependencies: []
+description: A KDE implementation in Ruby.
+email:
+- ryanmt@byu.net
+- logan@datacurrent.com
+executables: []
+extensions: []
+extra_rdoc_files: []
+files:
+- lib/kder.rb
+- lib/kder/bandwidth.rb
+- lib/kder/kde.rb
+- lib/kder/version.rb
+- lib/util/enum_extensions.rb
+- lib/util/statistics.rb
+homepage: https://github.com/loganb/KDER
+licenses: []
+metadata: {}
+post_install_message: 
+rdoc_options: []
+require_paths:
+- lib
+required_ruby_version: !ruby/object:Gem::Requirement
+  requirements:
+  - - ">="
+    - !ruby/object:Gem::Version
+      version: '0'
+required_rubygems_version: !ruby/object:Gem::Requirement
+  requirements:
+  - - ">="
+    - !ruby/object:Gem::Version
+      version: '0'
+requirements: []
+rubyforge_project: 
+rubygems_version: 2.5.2
+signing_key: 
+specification_version: 4
+summary: Kernel Density Estimation, and an associated bandwidth estimator, rudimentary
+  at best, but a decent first approximation for backend preparing of KDE plots for
+  plotting in JS libraries, or for whatever you would like to use it.
+test_files: []