Vaguery

require './stickpig'

100.times do |i|
  stick = StickpigScript.new(0.4)
  puts "#{i}, #{stick.script.length}" +
    [-20,-10,-1,0,1,10,20].inject("") {|line, x| line + ", #{stick.evaluate(x).round(5)}"}
end

Vaguery

class Float
  alias :add :+
  alias :subtract :-
  alias :multiply :*
  alias :negate :-@
  def divide(arg2)
    arg2==0 ? 0.0 : self / arg2
  end
end

Vaguery

Vaguery

def score(a=0,b=1,c=2,d=3,e=4)
  puts "a: #{a.class}"
  puts "b: #{b.class}"
  puts "c: #{c.class}"
  puts "d: #{d.class}"
  puts "e: #{e.class}"
  
  scores = Array.new
  scores << a << b << c << d << e

Vaguery

Feature: Pile interpreter handles tokens as expected
  In order to get consistent behavior when executing Pile scripts
  As a symbolic regression modeler
  I want each of the six basic token types to bring about expected changes in the Pile interpreter stack

  Scenario: numerical constants are pushed onto the interpreter stack
    Given the Pile script "-91.21"
    When I execute it
    Then there should be 1 item on the stack

Vaguery

Feature: random Pile script generator
  In order to get sample arbitrary Pile solutions to a target problem
  As a symbolic regression modeler
  I want to be able to compose random Pile scripts, sampled from a set of tokens

  Scenario: sample available tokens uniformly
    When I generate 10000 random Pile scripts, each with 20 tokens
    And the available tokens are ["a", "b", "c"]
    Then the tokens should appear (on average) with equal frequency

Vaguery

Feature: Pile return values
  In order to interpret a Pile script as a function of x
  As a symbolic regression modeler
  I want a scalar value to be returned from any executed script

  Scenario: return value is the top stack item after execution
    Given the Pile script "1.2 3.4 +"
    When I execute it with x = 1
    Then the returned value should be 4.6

Vaguery

Feature: Symbolic regression scoring against x+6 dataset
  In order to evaluate how well a Pile script models a given dataset
  As a symbolic regression modeler
  I want the score of a given script against a dataset to be the RSS value
  
  Scenario: score for a perfect program should be 0.0
    Given the Pile script "x 6.0 +"
    And the training data is "x_plus_6.csv"
    When I score the script
    Then the score should be 0.0

Vaguery

Feature: One-point uneven crossover works as specified
  In order to combine "good ideas" that have arisen independently in different Pile scripts
  As a symbolic regression modeler
  I want crossover to snip scripts up and stick them back together

  Scenario: crossover conserves the number of tokens
    Given parent_1 is the Pile script "+ + + + + + + + + +"
    And parent_2 is the Pile script "- - - - - - - - - -"
    When I apply uneven one-point crossover to those parents
    Then the result is a collection of 2 Pile scripts

Vaguery

Feature: Uniform mutation works as specified
  In order to discover new beneficial variations of scripts that are already OK
  As a symbolic regression modeler
  I want uniform mutation to replace tokens in "parent" scripts with new random tokens

  
  Scenario: uniform mutation replaces tokens with uniform random sample of available tokens
    Given parent_1 is the Pile script "+ + + + + + + + + +"
    And the available tokens are ["a", "b", "c", "d", "e"]
    And the chance of mutation is p=1.0