Assignment 1: All About torch.Tensor

Probability distributions in Tensors

In this notebook, I will be showing you how to build Tensors in PyTorch using different continuous and discrete probability distributions.

Note that this is different than using probability distributions to evaluate a tensor (e.g. using Bernoulli(torch.tensor[1])) and create distributions based on those tensors.

torch.Tensor.cauchy_
torch.Tensor.uniform_
torch.Tensor.normal_
torch.Tensor.bernoulli_
torch.Tensor.geometric_

# Import torch and other required modules
import torch

Function 1 - torch.Tensor.cauchy_()

This function will fill a new tensor with numbers gathered from the Cauchy probability distribution. More information about this distribution can be found at https://mathworld.wolfram.com/CauchyDistribution.html

# Example 1 - working (change this)
t1= torch.Tensor(3,3).cauchy_(median = 0, sigma = 1, generator = None)
t1

tensor([[ -1.6426,   2.8174, -20.1251],
        [ -0.9105,   0.0428,  -1.4505],
        [ -0.5068,   0.2860,  -0.4843]])

This example is generated using the three default arguments for this function, taking in a median parameter, sigma being the half width at the half maximum point of the distribution, and a generator parameter