CI Chapter 3 by Pro.Chu part3

Replacement 
 two class of GA, based on replacement 

• Generational genetic algorithm(GGA)

• Steady-state genetic algorithms(SSGA) this one is better than
   Replace random
   Replace worst 
   Replace oldest  if the oldest is good than the new one that not good
   Elitist never be replace because we mark this one is the best it's depend on your parameter (normally n=1) if we set more this test will not be vary in the fitness

the amount off overlap between current and new populations is referred to as the generation gap.
=============   all of those part are end of one generation or  iteration ===============

so next one is Stopping Criteria
there were 2 approaches for stopping criteria:
 - First approach
    Stop aafter production of definite number of generations (MAXGEN) or number of objective function calls evaluations (MAXNFC/MAXNFE<use this more>) or (rarely used) time (in s/min/hr)

- Second approach
    Stop when the improvement in average fitness over two generation is below is below a threshold

GA vs Ad-hoc

In disadvantage
   there are two words we should know that is exploration (diversification) การควานหาคำตอบบริเวณกว้าง  and exploitation (intensifation) การหาคำตอบบริเวณที่เฉพาะเจาะจง

effects of control parameters

Population size (PS or NP) - typical 50 - 200
Crossover rate (CR or Pc) - typical 0.70 - 0.90

if crossover rate is too high the diverse is low
if crossover rate is too low the search may stagnate due to the lower exploration rate

Mutation rate (MR or Pm) - typical 0.01 - 0.05
 Mutation increase the variability of the population.

High level of mutant yield an essentially random search, with may lead to very slow convergence
Low level of mutant

So next we have class experiment

MySimpleDiary