A floating point “Bit” climber

• Randomly generate and evaluate a float string

• Compute ? and ? based on each gene’s probabilities (a gene is a group of bits, say 10)

• Until stopping criteria satisfied:

  • Select a single string position, i
  • Generate a mutation value as N(?, ?)
  • Add mutation value to string(i)
  • Evaluate the new string
  • If fitness is worse, undo the mutation
  • Else, recompute ? and ? for that gene
  • If count exceeds a threshhold, rerandomize the string

• 10x faster than other bit climbers tested (2000x faster than original solution)

Previous slide

Next slide

Back to first slide

View graphic version