FairML/wp1_case_plot.py at master · eustomaqua/FairML · 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
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
# coding: utf-8
#
# TARGET:
#   Oracle bounds regarding fairness for majority voting


import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

# from pyfair.granite.draw_graph import (
from pyfair.facil.draw_prelim import (
    DTY_PLT,  # PLT_LOCATION, PLT_AX_STYLE,
    _setup_config, _setup_figsize, _setup_figshow, _barh_kwargs)


# -----------------------
# Case study (results)


model_string = ["Full", "Unaware", "Fair K", "Fair add"]

mse = {'org': [0.7755605, 0.80419207, 0.86384302, 0.86296964],
       'qtb': [0.90229589, 0.80419207, 0.86393648, 0.859056]}
rmse = {'org': [0.88065913, 0.89676757, 0.92943156, 0.92896159],
        'qtb': [0.94989257, 0.89676757, 0.92948184, 0.92685274]}

att1 = {
    'DP': {'org': [0.004948, 0.052723, 0, 0.01735],
           'qtb': [0.051301, 0.052723, 0, 0.014767]},
    'EO': {'org': [0.011466, 0.017947, 0, 0.018456],
           'qtb': [0.031385, 0.017947, 0, 0.013989]},
    'PQP': {'org': [0.034763, 0.016614, 0.041549, 0.040802],
            'qtb': [0.016532, 0.016614, 0.041549, 0.04199]},
    'DR': [0.4217655301, 0, 0, 0.01261092947],
    'loss': [0.3825315273, 0.4093881364, 0.4607659972, 0.466137319]}

att2 = {
    'DP': {'org': [0.775993, 0.347226, 0, 0.010956],
           'qtb': [0.247111, 0.347226, 0, 0.004904]},
    'EO': {'org': [0.738663, 0.321692, 0, 0.002868],
           'qtb': [0.26862, 0.321692, 0, 0.01135]},
    'PQP': {'org': [0.233629, 0.313737, 0.323618, 0.327041],
            'qtb': [0.348565, 0.313737, 0.323618, 0.325454]},
    'DR': [0.4217655301, 0, 0, 0.01261092947],
    'loss': [0.3825315273, 0.4093881364, 0.4607659972, 0.466137319]}


# -----------------------
# Case study plots


fair_string = ["DP", "EO", "PQP", "DR", "loss"]
bias_risk = ['undisturbed', 'disturbed']
bias_risk = ['unperturbed', 'perturbed']


def _subplot_bar_pl(ax, names, annot, val_1, val_2):
    num = len(names)
    ind = np.arange(num)  # x locations for group
    wid = .4  # width of the bars: can also be len(x) sequences
    h1_facecolor = tuple([0, 0.5019, 0.5020])
    h2_facecolor = tuple([i / 255 for i in [64, 224, 208]])

    # p1, p2 =
    ax.bar(ind - wid / 2, val_1, wid, color=h1_facecolor,
           label=bias_risk[0], **_barh_kwargs)
    ax.bar(ind + wid / 2, val_2, wid, color=h2_facecolor,
           label=bias_risk[1], linestyle='dashed', **_barh_kwargs)
    ax.legend(loc='lower right', labelspacing=.05)
    ax.set_ylabel(annot)
    ax.set_xticks(ind)
    ax.set_xticklabels(names, rotation=0)
    # ax.autoscale_view()
    return


def _subplot_scatter_pl(ax, ap, names, score, att, comparison=True):
    delta_score = np.subtract(score['org'], score['qtb'])
    delta_score = np.abs(delta_score)
    values = [np.subtract(att[i]['org'], att[i]['qtb']
                          ) for i in ['DP', 'EO', 'PQP']]
    val_alt = np.array(att['DR'])

    num = 2 + (3 if comparison else 0)
    ind = np.arange(4)  # x locations for group
    wid = .4  # width of the bars: can also be len(x) sequences
    colors = sns.color_palette(palette='muted', n_colors=5)

    if not comparison:
        # p0, p4 =
        ax.bar(ind - wid / 2, delta_score, wid,
               color=colors[0], label=ap,  # alpha=.5,
               linestyle='dashed', **_barh_kwargs)
        ax.bar(ind + wid / 2, val_alt, wid,
               color=colors[4], label='DR', **_barh_kwargs)
    else:
        wid = .8 / num
        # p0, p1, p2, p3, p4 =
        ax.bar(ind - 2 * wid, delta_score, wid,
               color=colors[0], label=ap,  # alpha=.5,
               linestyle='dashed', **_barh_kwargs)
        ax.bar(ind - wid, values[0], wid,
               color=colors[1], label='DP', **_barh_kwargs)
        ax.bar(ind, values[1], wid,
               color=colors[2], label='EO', **_barh_kwargs)
        ax.bar(ind + wid, values[2], wid,
               color=colors[3], label='PQP', **_barh_kwargs)
        ax.bar(ind + 2 * wid, val_alt, wid,
               color=colors[4], label='DR', **_barh_kwargs)

    ax.legend(loc='best', labelspacing=.05, frameon=False)
    # ax.set_ylabel('Fairness')
    ax.set_xticks(ind)
    ax.set_xticklabels(names, rotation=0)
    return


def counterfactual_fairness_case(score, att1, att2, annot, ap,
                                 figname='fair_cf', figsize='L-WS'):
    fig, ax = plt.subplots(figsize=_setup_config['L-NT'])

    _subplot_bar_pl(
        ax, model_string, annot, score['org'], score['qtb'])
    fig = _setup_figsize(fig, figsize, invt=False)
    _setup_figshow(fig, figname + '_model_score' + DTY_PLT)
    plt.clf()  # clear(fig)

    ax = fig.add_subplot(111)
    _subplot_scatter_pl(ax, ap, model_string, score, att1,
                        comparison=False)
    _setup_figsize(fig, figsize, invt=False)
    _setup_figshow(fig, figname + '_fair_nocmp' + DTY_PLT)
    # plt.show()
    return


if __name__ == '__main__':
    score = mse
    annot = 'MSE'
    ap = r'$\Delta(MSE)$'  # annot_prime
    counterfactual_fairness_case(score, att1, att2, annot, ap)
    # '''
    # score = rmse
    # annot, ap = 'RMSE', r'$\Delta(RMSE)$'
    # counterfactual_fairness_case(score, att1, att2, annot, ap)
    # '''


# Empirical results in manuscript
# """
# python wp1_case_plot.py
# """

# Empirical results not in use
# """
# python wp1_main_plot.py --draw -exp mCV_expt4 -dat german
# python wp1_main_plot.py --draw -exp mCV_expt4 -dat adult
# python wp1_main_plot.py --draw -exp mCV_expt4 -dat ppr
# python wp1_main_plot.py --draw -exp mCV_expt4 -dat ppvr
#
# python wp1_main_plot.py --draw -exp mCV_expt6 --nb-pru 7 -dat *ricci
# python wp1_main_plot.py --draw -exp mCV_expt4 -dat **german
# python wp1_main_plot.py --draw -exp mCV_expt5 --gather
# python wp1_main_plot.py --draw -exp mCV_expt5 -dat german
#
# python wp1_main_plot.py -exp mCV_expt11 --nb-cls 11 --name-ens *
# python wp1_main_plot.py -exp mCV_expt3 --nb-cls 21 --name-ens Bagging
# python wp1_main_plot.py -exp mCV_expt3 --nb-cls 11 --name-ens AdaBoostM1
# python wp1_main_plot.py -exp mCV_expt3 --nb-cls 11 --name-ens SAMME
# python wp1_main_plot.py -exp mCV_expt10 --name-ens Bagging --nb-iter 1
#
# python wp1_main_plot.py --draw -exp mCV_expt1 -dat **
# python wp1_main_plot.py --draw -exp mCV_expt2 -dat ** --ratio .5
# """