Multivariate linear regression in Ruby - adapted from example from Andrew Ng's Machine Learning Coursera class

linear_regressor.rb

require 'narray' # gem install narray

class LinearRegressor
  attr_reader :theta, :mean, :std, :cost_history
  
  def initialize opts = {}
    @alpha = opts[:alpha] || 0.01
    @iterations = opts[:iterations] || 400
  end

  def fit x, y
    x, @mean, @std = preprocess x
    y = NVector[*y]
    m = x.shape[1]
    @theta = NVector.float(x.shape[0])
    @cost_history = []
    cost = compute_cost x, y, m
    @cost_history << cost
    log cost, 0
    1.upto(@iterations) do |i|
      @theta -= (@alpha / m) * x.transpose * ((x * @theta) - y)
      cost = compute_cost x, y, m
      @cost_history << cost
      log cost, i
    end
    self
  end

  def predict x
    x, _m, _s = preprocess(x, @mean, @std)
    x * @theta
  end
  
  def fit_normal x, y
    x, @mean, @std = preprocess x
    y = NVector[*y]
    @theta = (x.transpose * x).inverse * (x.transpose * y)
    self
  end

  private

    def compute_cost x, y, m
      errors = (x * @theta) - y
      (1 / (2.0 * m)) * errors**2
    end

    def log c, i
      puts "[#{i}] err: #{c.to_f.round(4)}" if i % 10 == 0
    end

    def preprocess x, mean = nil, std = nil
      x = NMatrix.cast(x)
      x_mean = mean || x.mean(1)
      x_std = std || x.stddev(1)
      x_std[x_std.eq(0)] = 1.0 # so we don't divide by 0
      x = NMatrix.ref((NArray.ref(x) - x_mean) / x_std)
      out = add_ones_column x
      [out, x_std, x_mean]
    end

    def add_ones_column m
      out = NMatrix.float(m.shape[0] + 1, m.shape[1])
      out[1..m.shape[0], true] = m
      out[0, true] = 1
      out
    end
end


require 'csv'

x, y = [], []
CSV.read('ex1data2.txt').each do |row|
  x << row.slice(0,row.length-1).map(&:to_f)
  y << row.last.to_f
end

reg = LinearRegressor.new

reg.fit x, y

puts "Theta:"
p reg.theta

samples = [
  [2104,3],
  [1600,3],
  [2400,3]
]

puts "Predictions:"
p reg.predict samples


require 'gnuplot'

Gnuplot.open do |gp|
  Gnuplot::Plot.new( gp ) do |plot|
  
    plot.title  "Cost history"
    plot.xlabel "iteration"
    plot.ylabel "error"
    plot.terminal "png"
    plot.output "cost_history.png"
    
    x = 0.upto(400).to_a
    y = reg.cost_history

    plot.data << Gnuplot::DataSet.new( [x, y] ) do |ds|
      ds.with = "lines"
      ds.notitle
    end
  end
end

cost_history.png

ex1data2.txt

2104,3,399900
1600,3,329900
2400,3,369000
1416,2,232000
3000,4,539900
1985,4,299900
1534,3,314900
1427,3,198999
1380,3,212000
1494,3,242500
1940,4,239999
2000,3,347000
1890,3,329999
4478,5,699900
1268,3,259900
2300,4,449900
1320,2,299900
1236,3,199900
2609,4,499998
3031,4,599000
1767,3,252900
1888,2,255000
1604,3,242900
1962,4,259900
3890,3,573900
1100,3,249900
1458,3,464500
2526,3,469000
2200,3,475000
2637,3,299900
1839,2,349900
1000,1,169900
2040,4,314900
3137,3,579900
1811,4,285900
1437,3,249900
1239,3,229900
2132,4,345000
4215,4,549000
2162,4,287000
1664,2,368500
2238,3,329900
2567,4,314000
1200,3,299000
852,2,179900
1852,4,299900
1203,3,239500