def test_initial_guess3():
x0 = [-1, 1]
def f(args):
[x, y] = args
ans = 100 * (y - x**2)**2 + (1 - x)**2
return ans
demo = SteepestDescent(f, x0)
root_est = demo.find_root()
real_root = np.array([1., 1.])
assert real_root.all() == root_est.all()
def __init__(self):
self.current = Board()
self.best = Board()
random.seed()
self.k = 1
self.k_max = None
self.max_score = None
self.N = 0
self.local_search = SteepestDescent()
self.chances = {
Tile.CENTRAL: 0.0,
Tile.BORDER: 0.0,
Tile.CORNER: 0.0,
}
self.prob_map = {
Tile.CENTRAL: 0.0,
Tile.BORDER: 0.0,
Tile.CORNER: 0.0,
}
self.total = 25.0
def __init__(self):
self.current = Board()
self.best = Board()
random.seed()
self.k = 1
self.k_max = None
self.max_score = None
self.N = 0
self.local_search = SteepestDescent()
self.chances = {
Tile.CENTRAL : 0.0,
Tile.BORDER : 0.0,
Tile.CORNER : 0.0,
}
self.prob_map = {
Tile.CENTRAL : 0.0,
Tile.BORDER : 0.0,
Tile.CORNER : 0.0,
}
self.total = 25.0
from modeller import * from modeller.optimizers import actions from modeller.scripts import complete_pdb # Load our custom steepest descent optimizer from steepest_descent import SteepestDescent env = environ() env.io.atom_files_directory = ['../atom_files'] env.libs.topology.read(file='$(LIB)/top_heav.lib') env.libs.parameters.read(file='$(LIB)/par.lib') # Read in the initial structure: code = '1fdn' mdl = complete_pdb(env, code) atmsel = selection(mdl) # Generate the restraints: mdl.restraints.make(atmsel, restraint_type='stereo', spline_on_site=False) # Optimize with our custom optimizer: opt = SteepestDescent(max_iterations=80) opt.optimize(atmsel, actions=actions.trace(5))
class VNS:
def __init__(self):
self.current = Board()
self.best = Board()
random.seed()
self.k = 1
self.k_max = None
self.max_score = None
self.N = 0
self.local_search = SteepestDescent()
self.chances = {
Tile.CENTRAL: 0.0,
Tile.BORDER: 0.0,
Tile.CORNER: 0.0,
}
self.prob_map = {
Tile.CENTRAL: 0.0,
Tile.BORDER: 0.0,
Tile.CORNER: 0.0,
}
self.total = 25.0
def load_game_file(self, filename):
tiles = self.current.load_game_file(filename)
self.current.arrange(tiles)
self.N = self.current.N
self.max_score = 2 * self.N * (self.N - 1)
self.set_chances()
self.best = deepcopy(self.current)
self.k_max = self.N / 2
def set_chances(self):
curr_tot = 0.0
self.chances[Tile.CENTRAL] = 0.0
for i in range(1, self.k + 1, 1):
j = self.N - 2 * i
self.chances[Tile.CENTRAL] += 2 * j + 2 * (j - 2)
if self.k == self.k_max and self.N % 2 == 1:
self.chances[Tile.CENTRAL] += 1
self.chances[Tile.BORDER] = 4.0 * (self.N - 2)
self.chances[Tile.CORNER] = 4.0
for k, v in self.chances.items():
curr_tot += v
for k, v in self.chances.items():
self.chances[k] = self.chances[k] / curr_tot
def compute(self, sec=30):
print 'compute....'
current = self.best
timer = time.time() + sec
while (self.k == self.k_max
or time.time() < timer) and self.best.score < self.max_score:
self.generate_neighboor(current)
local = self.local_search.search(current)
if local.score > current.score:
current = deepcopy(local)
self.k = 1
else:
if self.k < self.k_max:
self.k = self.k + 1
self.set_chances()
if current.score > self.best.score:
self.best = deepcopy(current)
print 'best: {} '.format(self.best.score)
def generate_neighboor(self, board):
tile_list = self.generate_random_tile_type(board)
tile1 = self.generate_random_tile(tile_list)
tile2 = self.generate_random_tile(tile_list)
board.swap(tile1, tile2)
def generate_random_tile_type(self, board):
rnd = random.randint(0, 2)
t_chance = self.prob_map[rnd] / self.total
while t_chance > self.chances.get(rnd):
rnd = random.randint(0, 2)
t_chance = self.prob_map[rnd] / self.total
t = self.prob_map.get(rnd, None)
if t is not None:
self.prob_map[rnd] += 1
self.total += 1
return board.tiles[rnd]
def generate_random_tile(self, tiles_list):
rnd = random.choice(tiles_list)
while not self.isinneghiborhood(self.k, rnd.row, rnd.col):
rnd = random.choice(tiles_list)
return rnd
def isinneghiborhood(self, k, i, j):
expr = ( (i < k or i >= self.current.N-1-k) and 0 <= j <= self.current.N-1 ) or \
( (j < k or j >= self.current.N-1-k) and 0 <= i <= self.current.N-1 )
return expr
def print_exit_file(self):
with open('end.txt', 'w') as f:
for l in self.best.board_matrix:
for el in l:
u, r, d, l = el.tile.pieces
outstring = '{} {} {} {}\n'.format(u, r, d, l)
f.writelines(outstring)
class VNS:
def __init__(self):
self.current = Board()
self.best = Board()
random.seed()
self.k = 1
self.k_max = None
self.max_score = None
self.N = 0
self.local_search = SteepestDescent()
self.chances = {
Tile.CENTRAL : 0.0,
Tile.BORDER : 0.0,
Tile.CORNER : 0.0,
}
self.prob_map = {
Tile.CENTRAL : 0.0,
Tile.BORDER : 0.0,
Tile.CORNER : 0.0,
}
self.total = 25.0
def load_game_file(self,filename):
tiles = self.current.load_game_file(filename)
self.current.arrange(tiles)
self.N = self.current.N
self.max_score = 2*self.N*(self.N-1)
self.set_chances()
self.best = deepcopy(self.current)
self.k_max = self.N / 2
def set_chances(self):
curr_tot = 0.0
self.chances[Tile.CENTRAL] = 0.0
for i in range(1,self.k+1,1):
j = self.N -2*i
self.chances[Tile.CENTRAL] += 2*j + 2*(j-2)
if self.k == self.k_max and self.N % 2 == 1:
self.chances[Tile.CENTRAL] += 1
self.chances[Tile.BORDER] = 4.0*(self.N-2)
self.chances[Tile.CORNER] = 4.0
for k,v in self.chances.items():
curr_tot += v
for k,v in self.chances.items():
self.chances[k] = self.chances[k]/curr_tot
def compute(self,sec=30):
print 'compute....'
current = self.best
timer = time.time() + sec
while (self.k == self.k_max or time.time() < timer) and self.best.score < self.max_score:
self.generate_neighboor(current)
local = self.local_search.search(current)
if local.score > current.score:
current = deepcopy(local)
self.k = 1
else:
if self.k < self.k_max:
self.k = self.k + 1
self.set_chances()
if current.score > self.best.score:
self.best = deepcopy(current)
print 'best: {} '.format(self.best.score)
def generate_neighboor(self,board):
tile_list = self.generate_random_tile_type(board)
tile1 = self.generate_random_tile(tile_list)
tile2 = self.generate_random_tile(tile_list)
board.swap(tile1,tile2)
def generate_random_tile_type(self,board):
rnd = random.randint(0, 2)
t_chance = self.prob_map[rnd]/self.total
while t_chance > self.chances.get(rnd):
rnd = random.randint(0, 2)
t_chance = self.prob_map[rnd]/self.total
t = self.prob_map.get(rnd,None)
if t is not None:
self.prob_map[rnd] += 1
self.total += 1
return board.tiles[rnd]
def generate_random_tile(self,tiles_list):
rnd = random.choice(tiles_list)
while not self.isinneghiborhood(self.k,rnd.row,rnd.col):
rnd = random.choice(tiles_list)
return rnd
def isinneghiborhood(self,k,i,j):
expr = ( (i < k or i >= self.current.N-1-k) and 0 <= j <= self.current.N-1 ) or \
( (j < k or j >= self.current.N-1-k) and 0 <= i <= self.current.N-1 )
return expr
def print_exit_file(self):
with open('end.txt','w') as f:
for l in self.best.board_matrix:
for el in l:
u,r,d,l = el.tile.pieces
outstring = '{} {} {} {}\n'.format(u,r,d,l)
f.writelines(outstring)
import AnnoDomini.AutoDiff as AD
import numpy as np
from matplotlib import pyplot as plt
from scipy import linalg as la
from steepest_descent import SteepestDescent
from matplotlib import pyplot as plt
def f(args):
[x, y] = args
ans = 100 * (y - x**2)**2 + (1 - x)**2
return ans
x0 = [2, 1]
sd = SteepestDescent(f, x0)
root = sd.find_root()
traj = sd.xs
x = [traj[i][0] for i in range(len(traj))]
y = [traj[i][1] for i in range(len(traj))]
traj2 = np.concatenate(
(np.array(x).reshape(-1, 1), np.array(y).reshape(-1, 1)), axis=1)
# print(root)
# print(traj2)
def plot_descent(ans):
X, Y = np.meshgrid(np.linspace(-3, 3, 100), np.linspace(-2, 8, 100))
Z = f(np.array([X, Y]))
fig = plt.subplots(1, 1, figsize=(10, 7))
