🎲 Python Meets Excel

Coding in Excel Just Got Cooler

Sep 07, 2023

Here’s my take on what happened this week in the world of AI & why it matters 👀

If you enjoy this newsletter, please share this link and tag me so we can thank you. Find out more about us at diceytech.co.uk.

Thanks for reading Sofiane’s Snippets!

Hey everyone 👋🏿,

Hope you're having a fantastic week! I took a break for the past 2 weeks for life stuff (holidays, moving house…) but I’m back now!

🗞️ This week in AI → It was announced two weeks ago but I really think it’s important to mention, Excel gets supercharged with Python! Whether you're a Pythonista or an Excel wizard, your toolbox just got an awesome upgrade. As usual, with great power comes great responsibilities…
🧐 Learn → Want to know what techniques are actually working in practice on new datasets? Check out the State of Competitive Machine Learning report, where we highlight the top five insights from this year's edition.

Let's dive in 🌊

🗞️ This week in AI → Excel gets a Python boost!

👀 What’s new

Hey folks, guess what? Microsoft just dropped a bombshell! Python is coming to Excel, and it's in Public Preview as of now! No more juggling between Jupyter notebooks and spreadsheets; you can write Python code directly in Excel cells. Best part? It's running on the cloud, so you don't even have to worry about setup. 🤯

👉🏿 Introducing Python in Excel: The Best of Both Worlds for Data Analysis and Visualization (Blog)

🚨 Why it matters

Why is this a big deal, you ask? Well, Excel is like the bread and butter for millions of people who crunch numbers daily. Now, imagine adding some Python spice to that! You can do data cleaning, advanced visualizations, and even machine learning, all without leaving Excel. It's like turning your standard bicycle into a rocket-powered speed machine. 🚀

👉🏿 Open-source libraries and Python in Excel (Docs)

💡 Food for thought

Python in Excel is more than just a cool feature—it's a testament to the ever-growing importance of Python in the data world. So, what does this mean for you?

👉🏿 If you’re an excel wizz and you’ve been pushing back on learning python now is the time!

👉🏿 Is this an opportunity for a wave of "AI-powered" startups that are simply scikit learn in a spreadsheet

A word of caution though… An Excel workbook can now include regular functions, lambda functions, Excel scripts, macros, VBA, and Python code. However, this can make debugging and maintenance difficult. If the person who built the model leaves, expect the next person to have to start from scratch, potentially producing different results.

How team manages and collaborate around code inside an excel spreadsheet is still a question mark but I’m looking forward to see the workflows that emerge in the following months.

🧐 Learn → 5 Insights from The State of Competitive Machine Learning report (2022 edition)

Machine learning competitions are often a good indicator of what techniques actually work well in practice on new datasets.

The State of Competitive Machine Learning 2022 report contains interesting and surprising insights!

Transformers dominate natural language processing (NLP). ALL NLP-related winning solutions used transformers.
Pytorch is king for deep learning. Out of 46 winning solutions using deep learning, 44 used PyTorch, and only 2 used TensorFlow.
Convolutional Neural Networks (CNN) still dominate computer vision. And EfficientNet is the most popular pre-trained architecture for computer vision. Most people fine-tune pre-trained models rather than training from scratch.
Cross-Validation methods: Almost twice as many winning solutions used K-Fold Cross-Validation instead of a fixed validation set.
A big surprise for tabular competitions: the reign of XGBoost seems over. While gradient boosting still wins most tabular competitions, LightGBM is now the preferred approach, with CatBoost coming in second. XGBoost is third.

Psst.. Quick one! Enjoying this newsletter? Subscribe so you don’t miss out next week

🛠️ Cool Tools

Studdy AI (Product)

AI tutor in your pocket.

Studdy Step-by-Step breaks problems down into easy-to-understand steps. If you're confused by a step, just ask Studdy Buddy to explain it more.

Cambrian (Product)

Copilot for AI research

For researchers, scientists, and engineers in particular, staying up-to-date with the latest ML research is a real hassle. Cambrian was built to enable a more sophisticated deep search over 240,000 machine-learning papers while continually absorbing newly released work. It also enables rapid knowledge search through CambrianGPT, our in-house LLM for querying over individual papers.

🖼️ Meme of the week

💜 Enjoying this Newsletter?

Want to find your ideal career path through skills data? Start at diceytech.co.uk.

Discussion about this post

Ready for more?