CascadeMAP-Bayesian/optimize_pathways_example.py at master · loschmidt/CascadeMAP-Bayesian · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
from optimize_pathways_library import upload_content, generate_initial_ratios, generate_next_ratios
from artificial_fitness import simulate_values
import numpy as np
from pathlib import Path

if __name__ == '__main__':
    seed = None # fix for reproducibility, but note that the grid is generrated by Halton, so batches may converge to the same conditions if the seed is fixed it is recommended to set rand_if_same to True in this case)
    debug_flag = False # set to True for printing some intermediate output in the command line
    FILE_ARTIFICIAL_FITNESS = 'sym_data_artificial.txt' # name of the file with artificial fitness values to be used as simulated measurements

    FILE_INSTRUCTIONS = 'example-2d.json' # file with parameter setups (see other 'example-Xd.json' files for X=1, 2, 4, or 5)
    FILE_CONDITIONS = 'out_sorted.csv' # name of the file to save the conditions
    FILE_INIT_MEASUREMENTS = 'out_measured_init.csv' # name of the file to save the initial conditions and measurements
    N_INIT = 5 # number of conditions to generate initially

    FILE_ITER_CONDITIONS = 'out_next.csv' # name of the file to save new conditions generated for the current iteration
    FILE_ITER_MEASUREMENTS = 'out_measured_iter.csv' # name of the file to save conditions and measurements for current itreration
    N_ITER = 5 # number of iterations
    N_ITER_NEXT = 5 # number of conditions to generate per iteration
    BATCH_METHOD = 'LP' # 'KB' = Kriging Believer; 'LP' = Local Penalizer , or 'EE' = Exploration/Exploitation methods for batch generation
    BIAS_EXPLORE_EXPLOIT = 0 # value from -1 (primarily exploit) to 1 (primarily explore); 0 = default
    rand_if_same = False # if set to True, will enforce sampling different conditions per batch by replacing repetitions with random

    script_dir = Path(__file__).parent

    # uploading all the input data
    params, sim_groups, out_message = upload_content(FILE_INSTRUCTIONS)
    col_names = [item['name'] for item in params]
    if debug_flag:
        print(out_message)
        print(col_names)
        print(sim_groups)

    ### EXAMPLE: generate initial set for measurement
    output_sorted_file = FILE_CONDITIONS
    n_out = N_INIT # number of conditions to generate initially
    p_sorted = generate_initial_ratios(FILE_INSTRUCTIONS,
                                       n_out, output_sorted_file,
                                       debug_flag, seed,
                                       sort_output=True)

    if debug_flag:
        print(p_sorted)

    ### filling in simulated values
    file_in = FILE_CONDITIONS
    file_out = FILE_INIT_MEASUREMENTS
    file_train = FILE_ARTIFICIAL_FITNESS
    y_new = simulate_values(file_in, file_out, file_train, col_names)
    print("iteration %i, max = %d" % (0,np.max(y_new[:,0])))

    ### fitting a model and generating next set of point to test in a loop
    input_file = FILE_INIT_MEASUREMENTS
    iter_file = FILE_ITER_MEASUREMENTS
    output_file = FILE_ITER_CONDITIONS
    file_train = FILE_ARTIFICIAL_FITNESS
    n_out = N_ITER_NEXT
    n_iter = N_ITER
    max_iter = np.max(y_new[:,0])

    for i in range(n_iter):
        p_next = generate_next_ratios(FILE_INSTRUCTIONS, input_file,
                                      n_out, output_file,
                                      use_stdvs = False,
                                      debug_flag = debug_flag, seed = seed,
                                      batch_method = BATCH_METHOD,
                                      fixed_priority_param = False,
                                      rand_if_same = rand_if_same,
                                      explore_exploit_bias = BIAS_EXPLORE_EXPLOIT)
        if debug_flag:
            print(p_next)

        y_new = simulate_values(output_file, output_file, file_train, col_names)

        max_iter = max(max_iter, np.max(y_new[:,0]))
        print("iteration %i, max = %d, iter max = %d" % (i+1,max_iter,np.max(y_new[:,0])))

        data_old = np.loadtxt(input_file, skiprows=1)
        data_iter = np.loadtxt(output_file, skiprows=1)
        np.savetxt(iter_file, np.vstack((data_old,data_iter)),
                delimiter='\t', header='\t'.join(col_names), comments='')
        input_file = iter_file

    print(data_old)