Since my earlier post looking into SIMD vectors in .NET, I attempted a few more experiments, trying to understand better where they might be a good fit, and where they would not. The short version: at that point, my sense is that SIMD vectors can be very handy for some specific scenarios, but would require quite a bit of work to be usable in the way I was hoping to use them. This statement is by no means meant as a negative on SIMD; rather, it reflects my realization that a SIMD vector is quite different from a mathematical vector. All that follows should also be taken with a big grain of salt ..read more

Even though a lot of my work involves writing computation-heavy code, I have not been paying close attention to the System.Numerics namespace, mainly because I am lazy and working with plain old arrays of floats has been good enough for my purposes. This post is intended as a first dive into the question “should I care about .NET SIMD-accelerated types”. More specifically, I am interested in understanding better Vector<T>, and in where I should use it instead of basic arrays for vector operations, a staple of machine learning. Spoiler alert: some of my initial results surprised me, and I ..read more

The main reason I created Quipu is that I needed a Nelder-Mead solver for a real-world project. And, as I put Quipu through its paces on real-world data, I ran into some issues, revolving around “Not a Number” floating point values, aka NaN. tl;dr: the latest release of Quipu, version 0.2.2, available on nuget, should handle NaN values decently well, and has some minor performance improvements, too. In this post, I will go over some of the changes I made, and why. Fixing the main issue made me realize that I didn’t know floating point numbers as well as I thought, even though I have been usi ..read more

An old math problem I had not seen since my university days resurfaced the other day, the Maximum Flow problem. It came up in the context of analyzing some industrial process. For illustration purposes, let’s say we are producing sausages, following these steps: we grind some meat, add some seasoning, then stuff and tie the sausage casings, and split them into delicious sausage links. We could represent this process as a graph, like so: ``` mermaid graph TD; grind_meat --> season_meat; season_meat --> stuff_casing; stuff_casing --> cut_links; Now the question is, how m ..read more

In my previous post, I went over the recent changes I made to my F# Nelder-Mead solver, Quipu. In this post, I want to explore how I could go about handling constraints in Quipu. First, what do I mean by constraints? In its basic form, the solver takes a function, and attempts to find the set of inputs that minimizes that function. Lifting the example from the previous post, you may want to know what values of $(x,y)$ produce the smallest value for $f(x,y)=(x-10)^2+(y+5)^2$. The solution happens to be $(10,-5)$, and Quipu solves that without issues: #r "nuget: Quipu, 0.2.0" open Quipu.Nelder ..read more

Back in April ‘23, I needed a simple solver for function minimization, and published a basic F# Nelder-Mead solver implementation on NuGet. I won’t go over the algorithm itself, if you are curious I wrote a post breaking down how the Nelder-Mead algorithm works a while back. In a nutshell, the algorithm takes a function, and finds the set of inputs that produces the smallest output for that function. The algorithm is not foolproof, but it is very useful, and has the benefit of being fairly simple. After dog-fooding my library for a bit, I found some rough spots, and decided it was time to make ..read more

In September, I had the great pleasure of attending the Data Science in F# conference in Berlin. I gave a talk and a workshop on Linear Programming, and figured I would make the corresponding material available, in case anybody is interested: Presentation: An Ode to Linear Programming Workshop: 4 levels of Linear Programming Linear Programming is perhaps an unusual topic for a data science conference. It is certainly a departure from my usual focus on Machine Learning with F#! I wanted to talk about Linear Programming, and its extension, Mixed Integer Linear Programming, because in my view ..read more

This is a follow-up to my recent post trying to implement the classic Conway Game of Life in an MVU style with Avalonia.FuncUI. While I managed to get a version going pretty easily, the performance was not great. The visualization ran OK until around 100 x 100 cells, but started to degrade severely beyond that. After a bit of work, I am pleased to present an updated version, which runs through a 200 x 200 cells visualization pretty smoothly: As a side note, I wanted to point out that the size change is significative. Increasing the grid size from 100 to 200 means that for every frame, the nu ..read more

A couple of days ago, I came across a toot from Khalid Abuhakmeh, showcasing a C# + MVVM implementation of the Game of Life on Avalonia. I have been experimenting with Avalonia funcUI recently, and thought a conversion would be both a fun week-end exercise, and an interesting way to take a look at performance. Long story short, I took a look at his repository as a starting point, and proceeded to rewrite it in an Elmish style, shamelessly lifting the core from his code. The good news is, it did not take a lot of time to get it running, the less good news is, my version has clear performance is ..read more

In the recent weeks, I came across a use case which sounded like a good fit for a desktop application, which got me curious about the state of affairs for .NET desktop clients these days. And, as I was looking into this, I quickly came across Avalonia, and specifically Avalonia.FuncUI. Cross platform XAML apps, using F# and the Elmish loop? My curiosity was piqued, and I figured it was worth giving it a try. In this post, I will go over my first steps trying the library out. My ambitions are limited: first, how hard is it to get something running? Then, how hard is it to take an existing Avalo ..read more