make_mask.py

import numpy as np
from numpy.random import randint
import random
import math


def get_mask(view_num, data_size, missing_ratio):
    """
    :param view_num: number of views
    :param data_size: size of data
    :param missing_ratio: missing ratio
    :return: mask matrix
    """
    assert view_num >= 2
    miss_sample_num = math.floor(data_size*missing_ratio)
    data_ind = [i for i in range(data_size)]
    random.shuffle(data_ind)
    miss_ind = data_ind[:miss_sample_num]
    mask = np.ones([data_size, view_num])
    for j in range(miss_sample_num):
        while True:
            rand_v = np.random.rand(view_num)
            v_threshold = np.random.rand(1)
            observed_ind = (rand_v >= v_threshold)
            ind_ = ~observed_ind
            rand_v[observed_ind] = 1
            rand_v[ind_] = 0
            if np.sum(rand_v) > 0 and np.sum(rand_v) < view_num:
                break
        mask[miss_ind[j]] = rand_v

    return mask