Linklog | A Tale of Dynamic Programming
A very rich blog post on dynamic programming, very pedagogical too!
Nicolas Chagnet's Homepage
Welcome to my resume, portfolio and blog!
Latest posts
Linklog | TV Explorer
A bit different from the usual content I share, but this is webapp provides the best TV experience I've ever seen online. It's simple, doesn't require elaborate signups or setup, it relies on perfectly legal iptv/web sources and is incredibly responsive.
Linklog | Optimal Peanut Butter and Banana Sandwiches
There is a fine line between dedication and obsessiln, and this post toys with it. I love it!
Linklog | The quadratic sandwich
This post is a very well written explanation on whar mathematical conditions a loss function should obey so that gradient descent is effective on it.
Linklog | The Typestate Pattern in Rust
The typestate pattern is a very powerful way to handle state within the type system in Rust. This article is a great introduction to it!
Linklog | hotpath
If you need to profile some Rust crates, this might make your life easier.
Linklog | 👋 Hi. I'm Sam.
This blog has really cool interactive visualizations of complex concepts like weighted random sampling or load balancing.
Linklog | Semicolony
This website is a treasure trove of gorgeous, in-depth articles and interactive visualizations of systems engineering.
Linklog | Jujutsu megamerges for fun and profit
Yet another great article about jujutsu, this time it's about the megamerge workflow.
Blog | Owning your model deployment
A guide for data scientists on how to handle the deployment of your models, and how to maintain them in production.
Linklog | Programming Still Sucks.
What a poignant read. Very well written, hilarious at times, while carrying a real message.
Linklog | Datatype
With this variable font, you can make inline charts (rough but simple) from raw data.
Linklog | Flexoki
A beautiful color palette with a realistic paper-like feel to it.
Blog | Working with CSS a decade later
Rediscovering frontend work in a world of React and Tailwind.
Blog | Releasing `@nchagnet/remark-uv`
A quick project for a long weekend.
Linklog | How The Heck Does Shazam Work? (An Interactive Exploration)
If you've ever wondered how Shazam-like programs work, then read this. It's a wonderfully written and interactive explanation of these algorithms.
Linklog | Building Semantic Version Control in Rust
Stumbled upon this rather promising diff and merging tool (which can be integrated within git) to do semantic merges. Can't wait to try it!
Linklog | fnox
An interesting secrets manager from one of the devs behind mise.
Linklog | How a Kalman filter works, in pictures
A very cool (and thorough) explanation of Kalman filters with great visuals.
Blog | What does a typical person look like?
Due to the curse of dimensionality, a person perfectly average is quite rare, so what does a typical person look like?
Blog | Moving domain to nchagnet.eu
Finally I have my own space!
Blog | A brief history of entropy
A lengthy introduction to the concept of entropy in physics, from thermodynamics to statistical physics and quantum mechanics.
Linklog | Perfectly Normal
I like the "surprise" interpretation of the KL divergence presented in this post. It's quite a clear way to view the KL divergence.
Blog | Entropy as a measure of surprise
This post describes how the concept of entropy in statistics can be related to a well-defined notion of surprise, and how this can make entropy more intuitive.
Linklog | Gaussian Processes
This is a really cool explanation of Gaussian Processes from a more theoretical perspective.
Linklog | Jacobi Fields in Machine Learning
There are all sorts of applications of differential geometry to ML, and this article does a great job of introducing some key concepts.
Linklog | Hamilton-Jacobi-Bellman Equation: Reinforcement Learning and Diffusion Models
Another deep connection between physics and "AI": this time it's reinforcement learning (perhaps the original "AI") expressed as a Hamiltonian flow.
Linklog | Joins are NOT Expensive!
A rather strong argument that Kimball's dimensional modelling is still relevant these days and provides both storage savings and more efficient queries compared to OBT.
Linklog | Tis the Season...to be Bayesian!
Modeling holiday effects in timeseries forecasting is no easy feat, and this article shows how bayesian methods can really help with that. A great read!
Linklog | What Category Theory Teaches Us About DataFrames
An interesting perspective on the abstraction behind the dataframe algebra.
Linklog | MLU Explain
This website contains very high quality visualizations of common machine learning concepts, presented in a very pedagogical manner.
Linklog | Dijkstra's Shortest-Path Algorithm
A really great explanation of the Dijkstra algorithm with clear visualizations.
Blog | Random weighted sampling in a hurry
How to choose a random element using weighted sampling.
Blog | Bayesian statistics for confused data scientists
A gentle introduction to Bayesian statistics for data scientists who, like me, are confused by it.
Linklog | Neural Networks with Candle
Candle is a very interesting pytorch alternative in Rust. This article provides a simple introduction to this crate!
Linklog | When Data Lies: Finding Optimal Strategies for Penalty Kicks with Game Theory
This article highlights a very interesting aspect of analysing historical data. In this case, center penalty kicks appear surprisingly successful. What this article shows is that one shouldn't blindly assume that it's because they are inherently more effective (they aren't), it's because goalkeepers act suboptimally and the player-goalkeeper interaction is responsible for this overrepresentation.
Linklog | varlock
I've been on a "declarative" kick recently, and this sounds a really nice way to do this for environment files.
Linklog | Ten years late to the dbt party (DuckDB edition)
If like me you feel late to the dbt party, we're not alone!
Linklog | Reinforcement Learning on Operations Research Problem
This is a very interesting dive into practical applications of reinforcement learning to optimization problems.
Linklog | Learning Reinforcement Learning
And this is a thorough list of resources to learn more about RL.
Linklog | How I Structure My Data Pipelines
Linklog | How to Store Dotfiles - A Bare Git Repository
I've looked far and wide for a goot dotfiles management system, but a bare git repo is just the easiest and it's already present on every system!
Linklog | Feature Selection: A Primer
This is a very thorough review on the statistics behind the various measures used to select features when building a predictive model. Very nicely written!
Linklog | A Gentle Introduction to Graph Neural Networks
This is possibly the most complete reference on graphs and GNN I have seen. Moreover, the quality of the publication is incredible, with interactive visuals sprinkled through.
Blog | The reality of data science
A foray into what real data science looks like.
Linklog | SIMD programming in pure Rust
I always found SIMD quite confusing to apply and generally just cross my fingers that auto-vectorization will just do it for me. This article clarified a lot of things for me about this topic.
Linklog | Why DuckDB is my first choice for data processing
This post showcases all the reasons why I love and use DuckDB.
Linklog | Geometric Hallucination Detection via Directional Consistency in Embedding Space
This paper brings a very geometric approach to LLM hallucinations evaluation. It argues that using angles of displacement vectors between question and answer embeddings provides a powerful metric for hallucinations. Relevancy becomes a mean domain-centric vector (which can be calibrated beforehand) and deviations from that vector can be recognized as hallucinations.
Linklog | Why would I use DuckDB for that? | Robin's blog
A practical comparison between Postgres and DuckDB, if you're curious to see where each should be used.
Blog | DuckDB as a unified data interface
A description of how DuckDB has slowly become my unified interface to data everywhere.