Exercises and projects given in the Deep Learning ND program offered by Udacity.com.
The following topics will be addressed:
- the concept of tensors which is the main data structure of PyTorch and how tensors interact with NumPy Python package.
- the PyTorch module named autograd to calculate gradients for training neural networks. It is able to perform backpropagation by calculating the gradients at each operation in the network
- we will build a neural network and do the forward propagation
- we will define a loss and an optimization method to train a neural neywork (on a dataset of handwritten digits)
- we will understand how to test how a neural network is able to generalize using the validation procedure
- we will see that your neural network is not able to work well with more complex images, so we will learn a technique named transfer learning: to use pre-trained networks to improve the performances of your classifier