Neural Networks Basics

Before you start to write code

We recommend reading the Stanford University CS231n Course which is a good source of knowledge on neural networks. We could select a sub-list of important courses but they are actually all useful.

Code

Knowing how to use python/numpy properly is a must!

Pytorch

We all use Pytorch as it is the most flexible and intuitive deep learning framework. This power is mostly enabled by its capacity to build its compute graph on-the-fly. We strongly recommend you to learn about this framework!

Installing

Installing pytorch is easy, all you need is to follow this link.

Tutorials

Some useful resources on how to use the framework:

• Official Pytorch Tutorials;
• Code examples by Yannick Hold-Geoffroy;
• Train on MNIST step-by-step.

Useful Tools

Some tools that you can use to train your models:

• Remove your train loop boilerplate code with poutyne from the GRAAL team;
• Keep an orderly logs of your experiments with Comet.ml
• Visualize your training information with TensorboardX.