Why PyTorch?

Learn why Pytorch is the best choice for learning about GANs.

Building larger networks could be a laborious task. One of the most laborious parts is doing calculus to work out the relationship between the back-propagated errora technique for supervised learning of artificial neural networks using gradient descent algorithm. and weightsThe strength of connection between the network units. in our network. If we decide to change our network we would potentially have to do all this work again.

Here we’ll be using PyTorch because it takes away a lot of that leg work, and lets us focus on thinking about the design of our networks.

Relationship between backpropagation and weights in the network

One of the most powerful and convenient features of PyTorch is that it does all the calculus for us, whatever the shape or size of the network we dream up. If we change our minds about the design of the network, PyTorch automatically works out the new calculus without us having to pull out a pencil and paper to work out the gradients again.

PyTorch also tries really hard to look and feel just like normal Python. This means it’s easy to learn if you already know Python, and there are fewer surprises when using it.