A simple neural network example.

simple_neural_net.rb

class Neuron
  attr_accessor :inputs, :threshold, :weights

  def initialize(threshold, weights)
    @threshold = threshold
    @weights = weights
    @inputs = []
  end

  def output
    weighted = inputs.zip(weights).map do |v, w|
      # An input can be a number or another neuron. If it's a neuron, we need to get its output.
      v = v.output if v.respond_to? :output
      w * v
    end

    weighted.reduce(:+) >= threshold ? 1 : 0
  end
end

# Layer 1 takes input from 2 pairs of imaginary switches.
# We want the net to return 1 if only one of each pair is active.

# Layer 1
# This neuron acts as a NAND gate.
l1a = Neuron.new(0, [-1, -1])
# This neuron acts as an AND gate.
l1b = Neuron.new(2, [1, 1])
l1c = Neuron.new(0, [-1, -1])
l1d = Neuron.new(2, [1, 1])

# Layer 2
# These are both NAND gates. Plugging a NAND and an AND into another NAND creates an XOR.
l2a = Neuron.new(0, [-1, -1])
l2a.inputs = [l1a, l1b]
l2b = Neuron.new(0, [-1, -1])
l2b.inputs = [l1c, l1d]

# Layer 3
# This is an AND gate.
l3 = Neuron.new(2, [1, 1])
l3.inputs = [l2a, l2b]

[0, 1].product([0, 1], [0, 1], [0, 1]).each do |inputs|
  puts inputs.inspect
  pair1 = inputs.take 2
  pair2 = inputs.drop 2
  l1a.inputs = pair1
  l1b.inputs = pair1
  l1c.inputs = pair2
  l1d.inputs = pair2
  puts l3.output
end

# === Output ===
# [0, 0, 0, 0]
# 0
# [0, 0, 0, 1]
# 0
# [0, 0, 1, 0]
# 0
# [0, 0, 1, 1]
# 0
# [0, 1, 0, 0]
# 0
# [0, 1, 0, 1]
# 1
# [0, 1, 1, 0]
# 1
# [0, 1, 1, 1]
# 0
# [1, 0, 0, 0]
# 0
# [1, 0, 0, 1]
# 1
# [1, 0, 1, 0]
# 1
# [1, 0, 1, 1]
# 0
# [1, 1, 0, 0]
# 0
# [1, 1, 0, 1]
# 0
# [1, 1, 1, 0]
# 0
# [1, 1, 1, 1]
# 0