Understanding callbacks in fastai Mar 29, 2019 fastai is a great library for Deep Learning with many powerful features, which make it very easy to quickly build state of the art models, but also to tweak them as you wish. One of the best features of fastai is its callbacks system that lets you customize simply pretty much everything. However, it can take getting used to and that’s the purpose of this post: presenting the callback system in fastai, explaining how it works and how to use it and finally showing you a few examples. .... Read more ➦
How to initialize deep neural networks? Xavier and Kaiming initialization Mar 22, 2019 Initialization of neural networks isn’t something we think a lot about nowadays. It’s all hidden behind the different Deep Learning frameworks we use, like TensorFlow or PyTorch. However, it’s at the heart of why and how we can make neural networks as deep as they are today, and it was a significant bottleneck just a few years ago. In this post, I’ll walk over the initialization part of two very significant papers: .... Read more ➦
Finally understanding decorators in Python Mar 19, 2019 Python is a very beginner-friendly language. However, it also has some more advanced features that can be hard to grasp. For me at least, decorators were one of them. I was seeing them here and there but never fully understood what they did and how they worked, even after reading a bunch of stuff about them. Until I came across the talk ‘So you want to be a Python expert’ by James Powell in which he explains four key features of Python, including decorators. .... Read more ➦
Understanding precision and recall Feb 14, 2019 Precision, recall and the derived concepts are very important in Machine Learning (and more generally in statistics) yet can be hard to grasp. In this article I explain what they are, why they are useful and how to use them. What’s the point ? First off, it’s important to understand why those concepts were introduced and why they are useful. I am going to use an example throughout this post to make things clearer. .... Read more ➦
Setting Up a First Personal Website Jan 19, 2019 Having a personal website is very useful, for many reasons. However, building one is sometimes a bit difficult. Of course there’s WordPress, but it can feel a bit heavy, and too abstract for someone who likes to code. On the other hand, Web Development is not easy and can be time consuming. Here I’ll describe how I went about making my personal website (this one) and how I deployed it on GitHub Pages. .... Read more ➦
Roadmap to ML Nov 21, 2018 This is my roadmap to go from an undergraduate CS and math background to a very good understanding of Machine Learning and Deep Learning concepts, and more importantly to successfully apply ML and DL. I will put here everything I’ve done (in italic), I’m doing (in bold), and I’m planning on doing (normal text) in order to reach that goal. Courses Andrew Ng’s Coursera Machine Learning course fast.ai course part 1 v3 Udacity Intro to deep learning with PyTorch fast. .... Read more ➦