How to split odd number samples in an array?

Question

I have an array of values with a length of y (y = 7267). I am splitting the data based on x (x = 24) as shown below. I miss some values here because 7267/24 gives 302, not 302.8. This is because I am taking integer values. If I set int to float in the 3rd line, I get an error TypeError: 'float' object cannot be interpreted as an integer. How can I run the following code without losing any values in y? or maybe there is a better way of splitting the data like here?

import numpy as np

y = np.random.rand(7267)
samples = len(y)
x = 24

trim = samples % x
subsequences = int(samples/x)
sequence_trimmed = y[:samples - trim]
sequence_trimmed.shape = (subsequences, time_steps, 1)

You can use the `//` operator to force integer division instead of casting it -- `samples // x` — Brendan Abel, Jul 14 '20 at 17:50
that still gives 302. so basically using either `int(samples/x)` or `samples//x` is the same, no? — ferrelwill, Jul 14 '20 at 17:52
That throws error `ValueError: cannot reshape array of size 7248 into shape (303,24,1)` — ferrelwill, Jul 14 '20 at 17:56

score 2 · Answer 1 · edited Jul 15 '20 at 23:43

2

Use

subsequences = samples // x

// is integer division so it'll return an integer exp.

var // x + var % x = var / x

edited Jul 15 '20 at 23:43

Ann Zen

25,080
7
31
51

answered Jul 14 '20 at 17:50

12ksins

299
1
11

score 2 · Answer 2 · edited Jul 15 '20 at 23:44

2

The modulo operation (%) is only defined for integers, that's why you cannot change it to float. An array also takes only (integer,integer) shapes. If I understand you correctly and you want to have an array that is shaped (x, z) for some z and that will definitely take all your data, how about introducing some zeros? So

z=samples//x+1
array=np.zeros(x*z)
array[:samples]=y
np.reshape(array,(x,z))

might do the trick?

edited Jul 15 '20 at 23:44

Ann Zen

25,080
7
31
51

answered Jul 14 '20 at 17:59

Dididombi

31
3

Your reasoning is correct. However, I get 7272 instead of 7267 in the output. – ferrelwill Jul 14 '20 at 18:38

score 0 · Answer 3 · answered Jul 14 '20 at 17:58

0

In [28]: import numpy as np
    ...:
    ...: y = list(range(7267))
    ...: samples = len(y)
    ...: x = 24
    ...:
    ...: trim = round(samples / x)
    ...: subsequences = int(samples/x)
    ...: sequence_trimmed = y[:samples - trim]
    ...:

answered Jul 14 '20 at 17:58

bigbounty

14,834
4
27
58

You miss the last line of the code. It throws this error `ValueError: cannot reshape array of size 6964 into shape (302,24,1)` – ferrelwill Jul 14 '20 at 18:11
I'm using list not numpy array – bigbounty Jul 14 '20 at 18:12
But I have an array. – ferrelwill Jul 14 '20 at 18:13

score 0 · Answer 4 · answered Jul 14 '20 at 18:06

0

This?

import numpy as np

y = np.random.rand(7267)
samples = len(y)
x = 24

trim = samples % x
subsequences = samples//x
sequence_trimmed = y[:samples - trim]
sequence_trimmed.shape = (subsequences, x)

answered Jul 14 '20 at 18:06

How to split odd number samples in an array?

4 Answers4