np.random.choice throwing ValueError when using non-integer probabilities with large arrays
This might be a silly question, but I'm dealing with I'm integrating two systems and I've been struggling with this for a few days now and could really use some help... I'm working with a `ValueError` when trying to use `np.random.choice` with a probability array that contains floats that do not sum to 1, especially when the array size is large. I've defined a large dataset and a corresponding probability distribution, but when I run the code, I get the following behavior message: ``` ValueError: probabilities do not sum to 1 ``` Here's a snippet of my code: ```python import numpy as np # Create a large array of choices choices = np.arange(10000) # Create a probability distribution that sums to less than 1 probabilities = np.random.rand(10000) * 0.1 # This will sum to less than 1 # Attempt to sample from the choices sampled = np.random.choice(choices, size=100, p=probabilities) ``` I've tried normalizing the `probabilities` array using `probabilities / np.sum(probabilities)`, but I noticed that this only works for smaller arrays. When I increase the array size, the normalization seems to introduce issues due to floating-point precision. Is there a way to handle larger arrays or a better approach to ensure that the probability sums to 1 without causing errors? I'm using NumPy version 1.24.0. Any suggestions would be greatly appreciated! Has anyone else encountered this? Has anyone else encountered this? What's the correct way to implement this?