hyperactive.search_space

Complexity

Total Complexity

11

Size/Duplication

Total Lines	53
Duplicated Lines	0 %

Importance

Changes

0

5 Methods

Rating	Name	Duplication	Size	Complexity
A	SearchSpace.get_random_pos_scalar()	0	5	1
A	SearchSpace.pos_space_limit()	0	7	2
A	SearchSpace.__init__()	0	14	4
A	SearchSpace.pos2para()	0	8	3
A	SearchSpace.get_random_pos()	0	5	1

# Author: Simon Blanke
# Email: simon.blanke@yahoo.com
# License: MIT License

import random
import numpy as np


class SearchSpace:
    def __init__(self, _core_, model_nr):
        self.search_space = _core_.search_config[list(_core_.search_config)[model_nr]]
        self.pos_space_limit()
        self.init_type = None

        self.para_names = list(self.search_space.keys())

        if _core_.init_config:
            self.init_para = _core_.init_config[list(_core_.init_config)[model_nr]]

            if list(self.init_para.keys())[0] == list(self.search_space.keys())[0]:
                self.init_type = "warm_start"
            elif list(self.init_para.keys())[0] == "scatter_init":
                self.init_type = "scatter_init"

    def pos_space_limit(self):
        dim = []

        for pos_key in self.search_space:
            dim.append(len(self.search_space[pos_key]) - 1)

        self.dim = np.array(dim)

    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).astype(int)

        return pos

    def get_random_pos_scalar(self, hyperpara_name):
        n_para_values = len(self.search_space[hyperpara_name])
        pos = random.randint(0, n_para_values - 1)

        return pos

    def pos2para(self, pos):
        if len(self.search_space.keys()) == pos.size:
            values_dict = {}
            for i, key in enumerate(self.search_space.keys()):
                pos_ = int(pos[i])
                values_dict[key] = list(self.search_space[key])[pos_]

            return values_dict