def optimize(solver, mapper, nodes, target='rosen', **kwds):
if target == 'rosen': # 3d-rosenbrock
ndim = 3
actual_coeffs = [1.0] * ndim
pprint = list
else: # 4th-order chebyshev
from poly import chebyshev4coeffs as actual_coeffs
ndim = len(actual_coeffs)
from mystic.math import poly1d as pprint
# number of trials
N = nodes
print("Number of trials: %s" % N)
print("===============")
# initial guess
import random
x0 = ([random.uniform(-100,100) for i in range(ndim)] for i in range(N))
# minimize the function
results = mapper(nodes).map(solver, x0)
# find the results with the lowest energy
from optimize_helper import best_results
solution = best_results(results)
print("===============")
print("Actual params:\n %s" % pprint(actual_coeffs))
print("Solved params:\n %s" % pprint(solution[0]))
print("Function value: %s" % solution[1])
print("Total function evals: %s" % solution[4])
return
def optimize(solver, mapper, nodes, target="rosen", **kwds):
if target == "rosen": # 3d-rosenbrock
# Rosenbrock function
from dejong import rosen as the_model
ndim = 3
actual_coeffs = [1.0] * ndim
pprint = list
else: # 4th-order chebyshev
# Chebyshev cost function
from poly import chebyshev4cost as the_model
from poly import chebyshev4coeffs as actual_coeffs
ndim = len(actual_coeffs)
from mystic.math import poly1d as pprint
# number of trials
N = nodes
print "Number of trials: %s" % N
print "==============="
# initial guess
import random
x0 = ([random.uniform(-100, 100) for i in xrange(ndim)] for i in xrange(N))
model = (the_model for i in xrange(N))
# minimize the function
results = mapper(nodes).map(the_solver, model, x0)
# find the results with the lowest energy
from optimize_helper import best_results
solution = best_results(results)
print "==============="
print "Actual params:\n %s" % pprint(actual_coeffs)
print "Solved params:\n %s" % pprint(solution[0])
print "Function value: %s" % solution[1]
print "Total function evals: %s" % solution[4]
return
def optimize(solver, nodes, target='rosen', **kwds):
if target == 'rosen': # 3d-rosenbrock
# Rosenbrock function
from dejong import rosen as the_model
ndim = 3
actual_coeffs = [1.0] * ndim
pprint = list
else: # 4th-order chebyshev
# Chebyshev cost function
from poly import chebyshev4cost as the_model
from poly import chebyshev4coeffs as actual_coeffs
ndim = len(actual_coeffs)
from mystic.math import poly1d as pprint
# number of trials
N = nodes
print "Number of trials: %s" % N
print "==============="
# initial guess
import random
x0 = ([random.uniform(-100,100) for i in xrange(ndim)] for i in xrange(N))
model = (the_model for i in xrange(N))
# minimize the function
from itertools import imap # itertools
results = imap(the_solver, model, x0, **kwds)
# find the results with the lowest energy
from optimize_helper import best_results
solution = best_results(results)
print "==============="
print "Actual params:\n %s" % pprint(actual_coeffs)
print "Solved params:\n %s" % pprint(solution[0])
print "Function value: %s" % solution[1]
print "Total function evals: %s" % solution[4]
return
