def gaussian(ind, Pm, std, x_min, x_max):
'''Gaussian mutation with rate of success Pm. If the mutation results in
genes outside the specified bounds, they are set equal to the nearest
bound.
Arguments:
ind {'individual' object} -- The individual.
Pm {float} -- The rate of mutation as a decimal probability.
std {float} -- Standard deviation for the gaussian function.
x_min {float} -- Minimum bound on gene values.
x_max {float} -- Maximum bound on gene values.
Returns:
'individual' object -- The mutated individual, or None.
'''
if ind:
values = []
for x in ind.values:
if r.random() < Pm:
values.append(x + np.random.normal(0, std))
else:
values.append(x)
#stay within bounds
if values[-1] < x_min:
values[-1] = x_min
if values[-1] > x_max:
values[-1] = x_max
return individual(values)
return None
def two_point(Parents, n, Pc):
'''Two point crossover. Same as single point except a second crossover
point is calculated where the genome goes back to the first parent.
Arguments:
Parents {2-tuple of class 'individual'} -- The parents.
n {int} -- The genome length
Pc {float} -- The rate of success as a decimal propbability.
Returns:
'individual' object -- The child, or None.
'''
if r.random() < Pc:
child_values = []
crosspoint1 = r.randint(0, n)
crosspoint2 = r.randint(crosspoint1, n)
for i in range(n):
if i < crosspoint1:
child_values.append(Parents[0].values[i])
elif i >= crosspoint1 and i < crosspoint2:
child_values.append(Parents[1].values[i])
else:
child_values.append(Parents[0].values[i])
return individual(child_values)
return None #regenerate parent pair
def single_run(fdistr='RWS',
fsamp='p_sample',
fcross='single_point',
fmut='uniform'):
'''Driver for a single run of the GA. Genetic manipulation methods can be
left blank or supplied for varying outcomes.
Arguments:
fdistr {str} -- RWS or distribution method if provided.
fsamp {str} -- Proportional or sampling method if provided.
fcross {str} -- Single-Point or crossover method if provided.
fmut {str} -- Uniform or mutation method if provided.
Returns:
'individual' object -- The best-of-run individual.
'''
#initialize
best_of_run = individual([0] * n) #start with worst
gen_count = 1
P = []
while len(P) < N:
P.append(individual(None, n))
while gen_count <= max_generations:
distr(P, fdistr) #calculate distribution
for i in P: #check if any are better than current BOR.
if i.f > best_of_run.f:
best_of_run = i
P_next = []
while len(P_next) < N: #build next generation
i = mut(cross((sample(P, fsamp), sample(P, fsamp)), fcross), fmut)
if i: #if not None
P_next.append(i)
gen_count += 1
P = P_next
print('Total evaluations for run:', ind.evals)
return best_of_run
def uniform(ind, Pm, x_min, x_max):
'''Uniform random mutation method.
Replaces gene values regardless of current value.
Arguments:
ind {'individual' object} -- The individual.
Pm {float} -- The rate of mutation as a decimal probability.
x_min {float} -- Minimum bound on gene values.
x_max {float} -- Maximum bound on gene values.
Returns:
'individual' object -- The mutated individual, or None.
'''
if ind:
values = []
for _ in ind.values:
values.append(r.uniform(x_min, x_max))
return individual(values)
return None
def arithmatic(Parents, n, Pc, weight):
'''Arithmatic crossover. Each gene is calculated as ax[i] + (1-a)y[i].
Arguments:
Parents {2-tuple of class 'individual'} -- The parents.
n {int} -- The genome length
Pc {float} -- The rate of success as a decimal propbability.
weight {float} -- The weight that the first parent's genes have.
0.5 will result in each parent contributing equally.
Returns:
'individual' object -- The child, or None.
'''
if r.random() < Pc:
child_values = []
for i in range(n):
child_values.append( (weight * Parents[0].values[i]) + \
((1-weight) * Parents[1].values[i]) )
return individual(child_values)
return None
def single_point(Parents, n, Pc):
'''Single point crossover with rate of success = Pc
Arguments:
Parents {2-tuple of class 'individual'} -- The parents.
n {int} -- The genome length
Pc {float} -- The rate of success as a decimal propbability.
Returns:
'individual' object -- The child, or None.
'''
if r.random() < Pc:
child_values = []
crosspoint = r.randint(0, n)
for i in range(n):
if i < crosspoint:
child_values.append(Parents[0].values[i])
else:
child_values.append(Parents[1].values[i])
return individual(child_values)
return None #regenerate parent pair