def prepare_deterministic(n, k, d, r, max_gd_steps, max_nfe):
# Find directory for that specific n, k, d
directories = listdir('data')
n_string = 'n=' + str(n) + '_'
directories = [s for s in directories if n_string in s]
k_string = 'k=' + str(k) + '_'
directories = [s for s in directories if k_string in s]
d_string = 'density=' + str(d) + '.'
directories = [s for s in directories if d_string in s]
# Raise error if none or multiple directories were found.
if len(directories) == 0:
raise PlatypusError('Directory with those n, k, d not found!')
if len(directories) != 1:
raise PlatypusError(
'Multiple directories with those n, k, d was found!')
directory = 'data/' + directories[0] + '/'
# Find the names of files that store the values of R matrices.
matrices = listdir(directory)
matrices = [s for s in matrices if 'R' in s]
# Raise error if no files 'R*.csv' were found
if len(matrices) == 0:
raise PlatypusError('Directory with those n, k, d not found!')
R = read_R(directory, matrices, n)
gd = GradientDescentOptimizerAdam(R)
local_search = AdamLocalSearch(gd, steps=max_gd_steps)
problem = TriFactorization(gd)
generator = ConstantKMatrixGenerator(len(matrices), n)
hypervolume_dynamics = HypervolumeDynamics(k)
algorithm = deterministic_algorithm(
problem=problem,
local_search=local_search,
generator=generator,
hypervolume_dynamics=hypervolume_dynamics)
termination_condition = MaxEvaluationsAndTreshold(max_nfe)
return algorithm, termination_condition
# Find directory for that specific n, k, d dir = '../data' directories = listdir(dir) n_string = 'n=' + str(n) + '_' directories = [s for s in directories if n_string in s] k_string = 'k=' + str(k) + '_' directories = [s for s in directories if k_string in s] d_string = 'density=' + str(d) + '.' directories = [s for s in directories if d_string in s] directory = dir + '/' + directories[0] + '/' # Find the names of files that store the values of R matrices. matrices = listdir(directory) matrices = [s for s in matrices if 'R' in s] R = read_R(directory, matrices, n) gd = GradientDescentOptimizerAdam(R) local_search = AdamLocalSearchProgress(gd, steps=gd_steps_part1) local_search1 = AdamLocalSearchProgress(gd, steps=gd_steps_part_2) generator = ConstantKMatrixGenerator(len(matrices), n, scale=0.1) problem = TriFactorization(gd) # Saved errors for diffrerent k's gd_step_1_results = dict() gd_step_2_results = dict() create_folder(dir_for_pictures) # def optimization for random init