for testing and deploying your application
for finding and fixing issues
for empowering human code reviews
# Author: Simon Blanke
# Email: [email protected]
# License: MIT License
import random
import numpy as np
class SearchSpace:
def __init__(self, _core_, model_nr):
self.search_config = _core_.search_config
self.warm_start = _core_.warm_start
self.scatter_init = _core_.scatter_init
self.model_nr = model_nr
self.memory = {}
def pos_space_limit(self):
dim = []
for pos_key in self.para_space:
dim.append(len(self.para_space[pos_key]) - 1)
self.dim = np.array(dim)
def create_searchspace(self):
"""
para_space = {}
for para_key in search_config_temp.keys():
for param_str in search_config_temp[para_key].keys():
new_param_str = para_key + "." + param_str
para_space[new_param_str] = search_config_temp[para_key][param_str]
self.para_space = self.search_config[list(self.search_config)[self.model_nr]]
self.pos_space_limit()
def get_random_pos(self):
pos_new = np.random.uniform(np.zeros(self.dim.shape), self.dim, self.dim.shape)
pos = np.rint(pos_new)
return pos
def get_random_pos_scalar(self, hyperpara_name):
n_para_values = len(self.para_space[hyperpara_name])
pos = random.randint(0, n_para_values - 1)
def pos2para(self, pos):
if len(self.para_space.keys()) == pos.size:
values_dict = {}
for i, key in enumerate(self.para_space.keys()):
pos_ = int(pos[i])
values_dict[key] = list(self.para_space[key])[pos_]
return values_dict
SimonBlanke /
Hyperactive
