Scaling or, mapping each entry of numpy array to another range of values

Question

I am trying to map reinforcement learning continuous action values range(-1.0,1.0) to real output.

Suppose I have numpy action array actions = np.array([-1., 0.2, -0.3, 0.5]). Values for the array may be actions.min = -1. and actions.max = 1..

Now,I need to map each entry to separate range for example

• actions[0] mapped to range(-0.4 , 1.54)
• actions[1] mapped to range(1.4 , 1.54)
• actions[2] mapped to range(-2.4 , 2.0)
• actions[3] mapped to range(-1.54 , 0.4)

So, Mapped action mapped_actions = np.array([-0.4, 1.484, -0.86, -0.085])

My current solution is as followes:

import numpy as np

actions = np.array([-1., 0.2, -0.3, 0.5])
mapped_low_high = np.array([[-0.4, 1.54], [1.4, 1.54],[-2.4, 2.], [-1.54, 0.4]])

mapped_actions = np.zeros_like(actions)

for i in range(actions.shape[0]):
    mapped_actions[i] = np.interp(actions[i], (-1., 1.), mapped_low_high[i])

Is there are a better way to map actions array to mapped_actions?

Answer 1

Looping with numpy arrays is slow. It is faster to write a vectorized function that can operate on the entire array at once. Since we know that the given array will always be between -1 and 1 and we have an array of the new range, the code is a simple function that maps from one range to the other. The general technique is to map the original data to [0, 1] and then map that to [a, b].

import numpy as np

actions = np.array([-1., 0.2, -0.3, 0.5])
mapped_low_high = np.array([[-0.4, 1.54],
                            [1.4, 1.54],
                            [-2.4, 2.],
                            [-1.54, 0.4]])


def remap(arr, mappings):
    out = arr.copy()
    a, b = mappings.T

    # remap out to [0,1]
    # assumes arr is [-1,1]
    out += 1.
    out /= 2.

    # remap out to [a,b]
    out *= b - a
    out += a
    return out


remapped = remap(actions, mapped_low_high)  # array([-0.4  ,  1.484, -0.86 , -0.085])