Why most models are small and LLMs are large PLAIOFFS24. Image by Author. When training supervised machine learning models, early stopping is a commonly used technique to mitigate overfitting. Early stopping involves monitoring a model’s performance on a validation set during training and stopping the training process once the model’s performance doesn’t improve on this held-out data. This technique helps save computation time and resources while ensuring that the model doesn’t learn the noise and irrelevant patterns in the training data, which could reduce its ability to generalize ..read more

Machine Learning on GCP: from Notebooks to Pipelines Notebooks are not enough for ML at scale Photo by Sylvain Mauroux on Unsplash All images, unless otherwise noted, are by the author Advocating for AI There is a misunderstanding (not to say fantasy) which keeps coming back in companies whenever it comes to AI and Machine Learning. People often misjudge the complexity and the skills needed to bring Machine Learning projects to production, either because they do not understand the job, or (even worse) because they think they understand it, whereas they don’t. Their fir ..read more

Deepmind and Isomorphic labs just published a new paper that applies new AI concepts and methods to create a new tool that promises to be again revolutionary Google’s AI subsidiaries DeepMind and Isomorphic Labs are making waves in the scientific community… again. This time, it is with the release of AlphaFold 3, a new AI model that predicts molecular structures with unprecedented accuracy and is not just limited to proteins like the successful (and truly game-changer in biology) AlphaFold 2. Indeed, AlphaFold 3 handles proteins and their complexes with DNA, RNA, ligands, ions, and more, promi ..read more

Can we discover new physics with KAN? Can a Neural Net Discover New Physics? (Generated with DALLE-2 by Author) Last week while we were at an AI & Physics conference (EuCAIFCon), a lot of the discussions were on Foundational models and whether it is possible to discover potentially new laws in Physics using AI, and lo and behold: recently the KAN paper¹ came out in arXiv discussing possibilities of discovering/rediscovering physics and mathematical models using neural net. I got some time over the weekend to go through parts of this fascinating paper and here we will take a ..read more

5 tips I would’ve hugely benefitted from as a beginner In August I’ll be coming up on my second year as a full time data scientist! (Technically my third year, but I was an intern my first year working in the field.) Photo by Campaign Creators on Unsplash I’ve learned a tremendous amount in the past few years. And although I did go to university for data science, which taught me a lot, there’s only so much college can prepare you for when it comes to entering the workforce. Nothing really beats just getting a ton of first hand experience. Being able to make mistakes on my own, f ..read more

A first step to Large Language Models Photo by okeykat on Unsplash I’ve been wanting to write a tutorial on Large Language Models (LLMs) for a while now and since then I’ve been thinking about how to write a series of beginner-friendly articles to understand and get started with LLMs. In this article, I will attempt to provide a beginner-friendly introduction to LLMs and explain the key concepts in a simple way and without digging further into the technical aspects. My hope is that after reading this article, you will feel more comfortable reading more advanced documentation on  ..read more

How to use Exploratory Data Analysis to drive information from time series data and enhance feature engineering using Python Photo by Ales Krivec on UnsplashIntroduction Time series analysis certainly represents one of the most widespread topics in the field of data science and machine learning: whether predicting financial events, energy consumption, product sales or stock market trends, this field has always been of great interest to businesses. Obviously, the great increase in data availability, combined with the constant progress in machine learning models, has made this topic ev ..read more

Advices from a Physics Master’s Graduate turned Data Scientist Source: DALL·E Hi there! I’ve often been asked about transitioning from physics to data science, data analysis, or machine learning, particularly by students and newcomers to the field. Considering that I get this question a lot, I thought it would be beneficial to share my experiences and insights on this topic. I hope you find this post helpful! My name is Sara, and I have a Master’s degree in Physics. Currently, I work as a Data Scientist at a global energy company. In this post, I aim to share my personal journey into ..read more

Photo by petr sidorov on UnsplashShould we aim to be one? In a world constantly embracing new ideas and capabilities, data scientists who use increasingly complex algorithms to feed in larger and larger datasets seem more mystical than ever in terms of their prediction process. When working as a data scientist whose main deliverables are product demand forecasts, I couldn’t help but wonder: Are data scientists just modern-day fortune tellers with mysterious but powerful magic that bless businesses with wise decisions? This post will take a peek at the glamorous yet comprehensive worl ..read more

A versatile and promising approach for the feature selection task Photo by Edu Grande on Unsplash Feature selection is a critical step in many machine learning pipelines. In practice, we generally have a wide range of variables available as predictors for our models, but only a few of them are related to our target. Feature selection consists of finding a reduced set of these features, mainly for: Improved generalization — using a reduced number of features minimizes the risk of overfitting. Better inference — by removing redundant features (for example, two features very correlated ..read more