1

Thank you in advance for taking a look at my post.

I have a 2d np.array called actions with shape (2,x) which contains ints

I have another 1d np.array keys with elements of the same type to the first dimension of actions: actions[0]. I want to remove from actions all array elements which are in keys. I tried diff = actions[:, not actions[0] == kids_keys]but it returns a 3d array of (1,2,x) shape.

How can I get a (2,x) diff array back ?

For example: 

actions = [[121122, 211122, 221122, ... 455544, 545544][0, 0.35, 0.75, ... 1, -0.25]]
keys = [211122 221122]
# The operation I am looking for:
actions - keys = [[121122, ... 455544, 545544][0, ... 1, -0.25]]

The error: The dimmentions of the diff array become (2,1,80) for some reason I dont know! The dimentions of the diff array become (2,1,80)

Improve this question
3
  • It would be helpful if you posted code for this, so people can see where you are in the process of trying to solve it. Sample input, and expected output would be helpful as well. Commented Jun 2, 2020 at 3:49
  • What's the shape of not actions[0] == kids_keys? Commented Jun 2, 2020 at 4:42
  • The duplicate is transposed, but you can either transpose your output or just operate on the right dimensions. Commented Jun 2, 2020 at 12:21

2 Answers 2

Reset to default
2

Use np.isin:

mask = np.isin(actions[0], keys, invert=True)
result = actions[:, mask]
Improve this answer
Sign up to request clarification or add additional context in comments.

3 Comments

its seems to work great! I would like to ask what ~ means and if there is another way to do this operation faster.
@AlexanderIthakis The ~ was to invert the result of np.isin (so only those not in keys are kept), but I changed it since, although inverting is a very fast operation, np.isin accepts an invert parameter to do the same more efficiently as part of the function call itself. About being faster, if I knew a better way to do it I'd post it :) np.isin is fairly fast, if all keys are always unique you can pass assume_unique=True. The masking is advanced indexing, which is relatively expensive.
Hey, I appreciate the detailed answered. Yes, my keys, are unique, I will definitely change it. Thanks again
0

The following will filter out columns of actions that have a first row entry in the set keys:

import numpy as np
x = 10
actions = np.random.randint(5, size=(2,x))
print(actions)
keys = np.array([1,2,3])
print(keys)
filtered_actions = actions[:,~np.sum([actions[0,:] == key for key in keys], dtype=bool, axis=0)]
print(filtered_actions)
Improve this answer

1 Comment

Thank you for your answer! For some reason when I try your code in a terminal it works as intended, but when I use it in my code, I get the same problem I got with my approach. It returns, a ( 2, 1, x) dimention array instead of (2, x). Can you please help me.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.