Matching elements of two arrays depending on sorted values (Python)

Question

I'm a bit stuck with this one and tried some stuff mentioned here but didn't help. Basically, I've got these arrays:

returns = np.array([0.01, 0.2, 0.05, -0.01, 0.3, -0.5])
weights = np.array([0.3, 0.25, 0.2, 0.15, 0.1, 0])

The idea is to create dynamic weighting in which highest weight matches with highest return. I tried sorting two arrays (or lists) depending on one another but it doesn't help (as mentioned here: Sorting list based on values from another list?), since it really depends on weight initialization.

Desired output would be (weights change their position depending on max/min and in between values of the returns array):

returns = np.array([0.01, 0.2, 0.05, -0.01, 0.3, -0.5])
weights = np.array([0.15, 0.25, 0.2, 0.1, 0.3, 0])

This generally would go row by row through some simple iteration but I cannot seem to crack the problem first. Sorting both separately and then matching ascending / descending order) is no good, since returns array has to be unsorted. Any ideas?

Sorry, don't get what you want to do. maybe edit it with an example input, example output and how weights play a role in it? — Patrick Artner
– Patrick Artner, Commented Nov 20, 2020 at 11:46

mathfux · Accepted Answer · 2020-11-20 12:42:23Z

2

It looks like you need to use inverse of np.argsort. For simplicity, it could be called ranks. So you can get ranks of returns like so:

_, ranks = np.unique(np.argsort(returns), return_index=True)
>>> ranks
array([2, 4, 3, 1, 5, 0], dtype=int64)

And a remaining part is:

>>> np.sort(weights)[ranks]
array([0.15, 0.25, 0.2 , 0.1 , 0.3 , 0.  ])

Alternative way

Following discussion on similar topic, I found that assignment of inverse argsort could be done also in this way:

a1 = np.argsort(returns)
a2 = np.argsort(weights)
weights[a1] = weights[a2] 
>>> weights
array([0.15, 0.25, 0.2 , 0.1 , 0.3 , 0.  ])

edited Nov 20, 2020 at 12:42

answered Nov 20, 2020 at 12:25

mathfux

5,9792 gold badges21 silver badges38 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Nimantha · Accepted Answer · 2021-10-28 16:14:31Z

1

Check this out:

import numpy as np
returns = np.array([0.01, 0.2, 0.05, -0.01, 0.3, -0.5])
weights = np.array([0.3, 0.25, 0.2, 0.15, 0.1, 0])

temp_dict = dict(zip(sorted(returns), sorted(weights)))

for index, el in enumerate(weights):
    weights[index] = temp_dict[returns[index]]

print(returns)
print(weights)

The output will be:

[ 0.01  0.2   0.05 -0.01  0.3  -0.5 ]
[0.15 0.25 0.2  0.1  0.3  0.  ]

edited Oct 28, 2021 at 16:14

Nimantha

6,5376 gold badges32 silver badges78 bronze badges

answered Nov 20, 2020 at 12:20

lorenzozane

1,2791 gold badge7 silver badges15 bronze badges

Comments

Nimantha · Accepted Answer · 2021-10-28 16:14:10Z

0

First we find the correct ordering in 'returns' and then we apply this order to the sorted version of 'weights'.

import numpy as np

returns = np.array([0.01, 0.2, 0.05, -0.01, 0.3, -0.5])
weights = np.array([0.3, 0.25, 0.2, 0.15, 0.1, 0])

correctOrderReturns = np.argsort(returns)[::-1] # [4 1 2 0 3 5] Indices of the correct order from max to min in "returns"
sortedWeights = np.sort(weights)[::-1] # [0.3  0.25 0.2  0.15 0.1  0.  ] Ordered list from max to min of weights

for i in range (len(weights)):
    weights[correctOrderReturns[i]] = sortedWeights[i]    

print (returns) #[ 0.01  0.2   0.05 -0.01  0.3  -0.5 ]
print (weights) #[0.15 0.25 0.2  0.1  0.3  0.  ]

This prints the desired output:

Returns: [ 0.01 0.2 0.05 -0.01 0.3 -0.5 ]

Weights: [0.15 0.25 0.2 0.1 0.3 0. ]

edited Oct 28, 2021 at 16:14

Nimantha

6,5376 gold badges32 silver badges78 bronze badges

answered Nov 20, 2020 at 12:21

rodvictor

3493 silver badges17 bronze badges

Collectives™ on Stack Overflow

Matching elements of two arrays depending on sorted values (Python)

3 Answers 3

Comments

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related