随机播放 numpy 数组而不拆分

Question

I have a 4D numpy array for input to a CNN similar to this generated data:

import numpy as np
X = np.random.rand(20, 1, 10, 4)

And the corresponding label (in a separate array) y is

y = [0,0,0,....0]

So that the first instance of my input looks like:

>>>X[0]
array([[[0.11529038, 0.56951377, 0.64859216, 0.53927201],
        [0.24599472, 0.99658675, 0.61760602, 0.23245005],
        [0.21688713, 0.87376011, 0.80853348, 0.95649564],
        [0.01096112, 0.36735236, 0.23917356, 0.06020551],
        [0.14795334, 0.31689876, 0.902638  , 0.95702681],
        [0.59684508, 0.53496984, 0.91312413, 0.17465782],
        [0.37409845, 0.51140496, 0.32453245, 0.59066936],
        [0.64259922, 0.6586773 , 0.13101008, 0.71666185],
        [0.59971516, 0.96920186, 0.8566649 , 0.37763693],
        [0.34957495, 0.88521399, 0.30383687, 0.23567811]]])

>>>len(y)
20

I would like to shuffle my data set before feeding into the.network but I cannot use sklearn train_test_split which split the data into the train-test. My dataset is already separated into train test but would like to shuffle before the model fit.

Answer 1

train_idx = np.random.permutation(train_set.shape[0])
train_x, train_y = train_x[train_idx], train_y[train_idx]

Similarly for val and test.

Answer 2

sklearn has a shuffle method

from sklearn.utils import shuffle
X, y = shuffle(X, y)