majority_vote_paradigm_shift/real_exps.py at main · amazon-science/majority_vote_paradigm_shift · 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
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
from utils import *
from iaa_api import *
from scipy import stats
from statsmodels.stats.multitest import multipletests
import argparse
from competitors.data_handling_competitors import seed_everything

def parse_args():
    parser = argparse.ArgumentParser()
    parser.add_argument('--seed_values', '-s', type=int, nargs='+', default=[42], help='Values of the seed.')
    return parser.parse_args()

if __name__ == '__main__':
    args = parse_args()
    if args.seed_values == [42]:
        seed = 42
        seed_everything(seed=seed)
        toloka_methods = ['Dawid-Skene', 'MACE', 'GLAD']
        other_methods = ['ebcc','la_one_pass', 'la_two_pass', 'MV', 'IWMV', 'BWA']
        all_results = {}
        all_p_values = {}
        data_list  = sorted([entry for entry in os.listdir('data') if os.path.isdir(os.path.join('data', entry))])

        for item in tqdm(data_list):
            print(f"Dataset: {item}")
            data = pd.read_csv(f'data/{item}/label.csv')
            real_labels = pd.read_csv(f'data/{item}/truth.csv')
            dataset_stat = str(np.round(compute_dataset_stats(real_labels),2))
            item_mod = item + ' - ' + dataset_stat
            all_results[item_mod] = {}

            oracle_T = generate_real_T(labels_to_list(real_labels), annotation_dict_to_list(data))
            oracle_results = list(oracle_MAP(annotation_dict_to_list(data), oracle_T, compute_class_distribution(real_labels)).values())
            all_results[item_mod]['Oracle MAP'] = {}
            all_results[item_mod]['Oracle MAP']['Result'] = round(compute_exact_predictions(real_labels, oracle_results), 3)

            modified_data = np.array(annotation_dict_to_list(data), dtype=object)
            iaa = InterAnnotatorAgreementAPI(modified_data)
            iaa._build_t_matrix()
            estimated_map = list(oracle_MAP(annotation_dict_to_list(data), iaa._t_hat, np.array(iaa._label_distribution)).values())
            all_results[item_mod]['Estimated MAP'] = {}
            all_results[item_mod]['Estimated MAP']['Result'] = round(compute_exact_predictions(real_labels, estimated_map), 3)
            if estimated_map != oracle_results:
                t_value, p_value = stats.wilcoxon(estimated_map, oracle_results)
            else:
                t_value, p_value = -1, -1
            all_results[item_mod]['Estimated MAP']['T value'] = round(t_value, 6)
            all_p_values['Estimated MAP'] = p_value

            if other_methods != []:
                e2wl, w2el, label_set = gete2wlandw2el(f'data/{item}/label.csv')
                for single_method in other_methods:
                    result = obtain_competitor_results(single_method, e2wl, w2el, label_set)
                    all_results[item_mod][single_method] = {}
                    all_results[item_mod][single_method]['Result'] = round(compute_exact_predictions(real_labels, result), 3)
                    t_value, p_value = stats.wilcoxon(result, oracle_results)
                    all_results[item_mod][single_method]['T value'] = round(t_value, 6)
                    all_p_values[single_method] = round(p_value, 6)
                    if single_method == 'MV':
                        all_results[item_mod][single_method]['Equality Percentage'] = count_number_occurrences(result, oracle_results, estimated_map)

            data = data.rename(columns={'item' : 'task'})

            if toloka_methods != [] :
                for method_name in toloka_methods:
                    method = obtain_toloka_method(method_name)
                    result = method.fit_predict(data)
                    all_results[item_mod][method_name] = {}
                    all_results[item_mod][method_name]['Result'] = round(compute_exact_predictions(real_labels, result), 3)
                    t_value, p_value = stats.wilcoxon(result.astype(float), oracle_results)
                    all_results[item_mod][method_name]['T value'] = round(t_value, 6)
                    all_p_values[method_name] = round(p_value, 6)

            rejected , corrected_p_values, _, _ = multipletests(list(all_p_values.values()), alpha=0.05, method='bonferroni')

            for (name, value), reject  in zip(all_p_values.items(), rejected):
                all_results[item_mod][name]['Stat Sig'] = reject

        result_df = flatten_results(all_results)
        result_df = result_df.round(decimals=4)
        os.makedirs('results', exist_ok=True)
        result_df.to_csv(f'results/real_data_results_{seed}.csv')
    else:
        all_results = {}
        toloka_methods = []#['Dawid-Skene', 'MACE', 'GLAD']
        other_methods = ['ebcc', 'BWA'] #['la_one_pass', 'la_two_pass', 'MV', 'IWMV', 'BWA']
        data_list  = sorted([entry for entry in os.listdir('data') if os.path.isdir(os.path.join('data', entry))])
        for seed in args.seed_values:
            seed_everything(seed=seed)
            all_results[seed] = {}
            for item in tqdm(data_list):
                print(f'Dataset: {item} - Seed: {seed}')
                data = pd.read_csv(f'data/{item}/label.csv')
                real_labels = pd.read_csv(f'data/{item}/truth.csv')
                dataset_stat = str(np.round(compute_dataset_stats(real_labels),2))
                item_mod = item + ' - ' + dataset_stat
                all_results[seed][item_mod] = {}

                oracle_T = generate_real_T(labels_to_list(real_labels), annotation_dict_to_list(data))
                oracle_results = list(oracle_MAP(annotation_dict_to_list(data), oracle_T, compute_class_distribution(real_labels)).values())
                all_results[seed][item_mod]['Oracle MAP'] = {}
                all_results[seed][item_mod]['Oracle MAP']['Result'] = round(compute_exact_predictions(real_labels, oracle_results), 3)

                modified_data = np.array(annotation_dict_to_list(data), dtype=object)
                iaa = InterAnnotatorAgreementAPI(modified_data)
                iaa._build_t_matrix()
                estimated_map = list(oracle_MAP(annotation_dict_to_list(data), iaa._t_hat, np.array(iaa._label_distribution)).values())
                all_results[seed][item_mod]['Estimated MAP'] = {}
                all_results[seed][item_mod]['Estimated MAP']['Result'] = round(compute_exact_predictions(real_labels, estimated_map), 3)

                if other_methods != []:
                    e2wl, w2el, label_set = gete2wlandw2el(f'data/{item}/label.csv')
                    for single_method in other_methods:
                        result = obtain_competitor_results(single_method, e2wl, w2el, label_set)
                        all_results[seed][item_mod][single_method] = {}
                        all_results[seed][item_mod][single_method]['Result'] = round(compute_exact_predictions(real_labels, result), 3)

                data = data.rename(columns={'item' : 'task'})

                if toloka_methods != [] :
                    for method_name in toloka_methods:
                        method = obtain_toloka_method(method_name)
                        result = method.fit_predict(data)
                        all_results[seed][item_mod][method_name] = {}
                        all_results[seed][item_mod][method_name]['Result'] = round(compute_exact_predictions(real_labels, result), 3)
        result_df = generate_table_with_std(all_results)
        result_df = result_df.round(decimals=4)
        os.makedirs('results', exist_ok=True)
        result_df.to_csv(f'results/real_data_results_multiple.csv')