Evolution vs Learning
   learning is optimizing a fitness function?

Black Box Optimization
  optimize function f:
    answer: f(x) > f(y)   is sufficient to do evolution

  function -> RANGE (output)  f(x)  x is in the DOMAIN (input)
                    phenotype                  genotype

  evolution:  searching through genotype space for good fitness

GLOBAL OPTIMUM:
find x such that fitness(f(x)) >= fitness(f(y)) forall y

we can never really know if we have found x.

f(x) :   x  --> f(x)

LOCAL OPTIMUM:
   must have a neighborhood!!!!!!
   neighborhood <-> mutation

function + neighborhood = landscape
landscape is required to have a local optimum

change the mutation -> changes the neighborhood -> changes the
landscape -> can change the number of local optima!!

REPRESENTATION:
100 bit string
99 neighbors

problem -> problem structure -> representation to use  -> mutation

exploration vs exploitation
    to make choices in mutation and representation

A RANDOM SEARCH

   init X
   while (fitness(X)<goodenough) {
       X' = mutate(X)
       if (fitness(X')>fitness(X)) X=X'
   }