.poisson()

The .poisson() function in PyTorch generates a tensor where each element is drawn from a Poisson distribution with a rate parameter provided in the input tensor. The function is useful for modeling count-based data, such as event occurrences over time.

Syntax

torch.poisson(input, generator=None)

• input: A tensor containing rate parameters for the Poisson distribution. Each value must be non-negative.
• generator (Optional): A random number generator for reproducibility. Defaults to None.

It returns a tensor with the same shape as input, where each element is a Poisson-distributed random value.

Example

The following example demonstrates how to use .poisson():

import torch

# Create a tensor containing Poisson rate parameters
rates = torch.tensor([2.0, 5.0, 10.0, 3.5])

# Generate samples from the Poisson distribution
samples = torch.poisson(rates)

# Print the resulting tensor
print(samples)

The code above generates the output as follows:

tensor([4., 6., 7., 6.])

Note: Since .poisson() generates random samples, the output values may vary each time the code is run.