Hi everyone!

Since I'm so involved in the machine learning and data science community I've built up quite a big pile of interesting project/resource/blog posts that is ever expending.

I'm also involved in tutoring data scientist and researchers these days, so I have quite a lot of effective tips to share!

Therefore, this year I'll be sharing a newsletter covering the most interesting of these resource I've found every week along with a useful data science tips!

Useful Resource

This week I'll be sharing 5 books on Machine Learning that I highly recommend. No particular order, I read all of them and they were truly helpful throughout my PhD! These are mostly all classic books that you should be aware of and use them efficiently to elevate your understanding of machine learning! The last two are more practical and hands-on, but I love how they showcase the content!

The Elements of Statistical Learning By Astie et al.

A classic in the machine learning community, highly recommend spending time with this book over and over again. I wouldn't start my machine learning journey with this, but this is the book to ensure your foundation is solid! A lot of questions can be cleared out throughout your machine learning career just by reading that book!

Artificial Intelligence, A Modern Approach (3rd Edition)

This will give you a good overview about the broader field of artificial intelligence which doesn't necessarily include machine learning. I would say this is the kind of book that enable you to start being creative with machine-learning and AI since it shows you all the scaffolding of the field!

The Deep Learning Book

Big classic in the deep learning field, it's surprisingly a very approachable book if you have some background in linear algebra (there is a primer at the beginning). I would go through it at least once in order to see the reasoning behind deep neural network straight from experts in the field; Ian Goodfellow, Yoshua Bengio and Aaron Courville.

Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems

If you had one book to go through to get up and running in the machine learning field, it would be this one. Very thorough and hands-on. I would go for this one before anything else if I were starting learning machine learning with no prior background.

Deep Learning with Pytorch

I love this book because you get a direct insight by the creators of Pytorch about their philosophy (and there are some neat neuroscience examples in there which I'm always happy to see). It explain beautifully not only the theory behind deep neural network, but also how to use Pytorch effectively!

btw if you have some cool projects or resource to share don't hesitate to reply to this email with a link to it!

Data Science Tips: Understand Why Your Project Exist

If you truly wants to make impactful data science work, no matter the field, one thing you have to master early on is the ability to understand why your project exist.

I would say that, above even technical skills, this is the most important ability of a data scientist.

Spending time understanding clearly the motives behind the project, even if you are the one starting it, has three major benefit:

It Gives you Enough Context to Modify it 🗒️

In order to creatively re-shape the project with confidence as you gather more information about the data you need to understand the context.

I've never seen a research / data science project that went exactly as planned upfront. Projects either change incrementally, with big pivots or fails.

It Focus You on What Matters 🎯

Understanding why your project is relevant allows you to focus your limited time on the parts that truly matters. This knowledge will allow you to spend more time on the area that are impactful and less on out-of-scope ones.

This is key, because there is more things to improve in a project than resource available. Starting by improving the parts that have the highest ROI and then going down the priority list is a much better strategy than just randomly improving any parts.

It Enable You to Cancel the Project 🙅

It's a bit awkward to say, but a lot of data science projects I've seen shouldn't even have been started in the first place (I've been part of some).

At least not in their initial state. Ensuring that whatever you are working on is impactful for your stakeholders (including you) is the best use of your time.

Conclusion

Talk to your stakeholders, take notes and understand deeply why your project exist before writing any lines of code. It's a bit of a controversial take, but don't touch that keyboard until you understand the reasoning and that you agree with it.

For machine-learning content checkout my Youtube channel covering these topics on a technical and theoretical level!

Have a wonderful week everyone! 👋

---

Extra: Generative 2D Art ✨

This week I've been tinkering about some generative art system that is using text prompt to generate 2D image. The system was made by Rivers Have Wings which is an artist outputting some next level contents on deep generative art. The system is using CLIP to guide a diffusion model neural network to generate images.

Here is my latest creation "The Great Yolk Falls from Heaven"

It's fairly small because I couldn't compute too much on Google Colab (I'll be setting up my GPU soon so I can do whatever I want on my own computer)!

The prompt I used is this segment Il n’y avait pas grand-chose à dire de plus dans cette pièce. by Ivan Aivazovsky taken from a french text I've written way back.

I'll be tinkering with this network a lot more and create a walk through of the code , it's super fun!