μβ

In a previous post, I detailed some of the features of MathOptSetDistances.jl and the evolution of the idea behind it.

Table of Contents MathOptInterface and the motivation Examples Set projections User-defined distance notions Bonus In this post, I will develop the process through which the MathOptSetDistances.jl package has been created and evolved.

Image source 1.

One of the luxuries of a PhD in applied maths / computer science is the possibility to work from home.

Table of Contents Automatic Differentiation Automatic differentiation on a pure-Julia solver Example problem: weighted independent set Optimization model of the weighted independent set A Julia implementation Why not reverse-mode?

Unlike other ecosystems in the scientific programming world, scientists and engineers working with Julia usually prefer a whole stack in Julia for many reasons.

A couple weeks ago, I had a wonderful evening thanks to the Skype a Scientist program, a 4th grade class from the US (think 9-10 years old if like me, you have no idea what grades stand for) and their super-dedicated teacher.

Constructors are a basic building block of object-oriented programming (OOP).

The progress of mathematical optimization as a domain has been tightly coupled with the development and improvement of computational methods and their implementations as computer programs.

In a previous post, we pushed the boundaries of the Graphs.jl abstraction to see how conforming the algorithms are to the declared interface, noticing some implied assumptions that were not stated.

In various graph-related algorithms, a graph is modified through successive operations, merging, creating and deleting vertices.

Last Friday was a great seminar of the Combinatorial Optimization group in Paris, celebrating the 85th birthday of Jack Edmonds, one of the founding researchers of combinatorial optimization, with the notable Blossom matching algorithm. .

This week, I came across Richard Oberdieck’s post, “Why ’evaluate’ is the feature I am missing the most from commercial MIP solvers”.

I must admit I am not always the most talented at social events.

This post explores the possibility to build static lists in Julia, meaning lists for which the size is known at compile-time.

While reading books, lots of thoughts come and go, especially if the subject resonates and is connected to topics of interest.

In a recent pull request on a personal project, I spent some time designing an intuitive API for a specific problem.

After the first submissions to journals, most researchers will be contacted by editors for reviewing articles others have written.

Enjoying the calm of the frozen eastern French countryside for the last week of 2018, I was struck by nostalgia while reading a SIAM news article [1] on a near-reversible heat exchange between two flows and decided to dust off my thermodynamics books (esp…

So, it’s been a bit more than a year since I took a flight to Montréal to start a PhD on mathematical optimization & game theory for smart grids.

This is an adapted post on the talk we gave with James at JuliaCon 2018 in London.

In the previous post, we explored a well-known integer optimization situation in manufacturing, the cutting stock problem.

Integer optimization often feels weird (at least to me).

With the end-of-year celebrations, we all had some expenses to manage, some of them shared with friends, and we all have this eternal problem of splitting them fairly.

In the last article, we explored different modeling options for a three-component systems which could represent the dynamics of a chemical reaction or a disease propagation in a population.

DifferentialEquations.jl came to be a key component of Julia’s scientific ecosystem.

The start of my journey as a PhD student last September was a big step, but also an opportunity to review and improve my working habits.

Last weeks have been pretty intense.

Two weeks ago, the Equisense team spent a couple days together near the ocean in northern France.

When I came back to Equisense, I was surprised and intrigued by many things.

In politics, a government’s first decisions and actions are often reviewed and assessed after the famous first 100 days.

It’s monday morning and I’m waking up with a hangover sensation, not from partying in the streets of Lille as one could think at the beginning of the week, but from the last intense 48 hours.

The goal of this article is to present couple challenges waiting the industrial data scientist or industrial data science teams, the deep reasons I believe are the root of this inertia, based on my experience (in both data science and engineering projects)

[1] Part III: Model development To follow the following article without any trouble, I would recommend to start with the beginning.

[1] Part II: Feature engineering What is feature engineering?

A complete predictive modeling project in Python Part I: Preprocessing and exploratory analysis One of the amazing things with data science is the ability to tackle complex problems involving hidden parallel phenomena interacting with each other, just from