March 2, 2026
Matrix mishaps & spicy sums
Notes on Lagrange Interpolating Polynomials
Math post sparks nitpick nuke, engineers flex, and a stability link drops jaws
TLDR: Eli explains how to fit a perfect curve through given points, and the comments immediately nitpick a matrix and a theorem typo while pros push real-world FEM and a stability-friendly barycentric method. Big picture: the math is solid, but the community insists technique and numerics decide whether it works in practice.
Eli Bendersky drops a clean explainer on how to build a curve through your data points, spotlighting the Vandermonde matrix and the Lagrange trick, and the comments instantly light up. The pedant patrol moves first: wolfi1 flags that the last matrix isn’t the identity, and commandersaki catches a letter swap in a theorem—q(x), not r(x). Meanwhile, the pros bring receipts. hdrz says Lagrange polynomials are the everyday workhorses inside Finite Element Method (FEM) software—think tools that simulate how bridges bend and heat flows—so this isn’t just textbook math, it’s production-grade. Then wenc arrives with a chemical engineering flex: turning tough differential equations into algebra using Orthogonal Collocation on Finite Elements (OCFE), using special “Radau” points—translation: pick smart checkpoints, get accurate results. Finally, a veteran cautions against the “just invert the matrix” vibe: TimorousBestie links the classic Trefethen–Boyd piece on the barycentric method, warning that naive interpolation can wobble numerically. The mood? Half textbook police, half real‑world war stories, with a side of humble pie for anyone who thought this was plug‑and‑play. Jokes about an “identity crisis” and “polynomial drama” fly, but beneath the memes, the takeaway is clear: the curve exists and is unique, but how you compute it really matters.
Key Points
- •Polynomial interpolation seeks a polynomial that passes through a set of distinct points.
- •Substituting data into a general polynomial yields a linear system with a Vandermonde matrix, which is invertible, proving existence and uniqueness.
- •Vandermonde matrices are often numerically ill‑conditioned, so direct inversion is not recommended in practice.
- •Lagrange basis functions are constructed to be 1 at one node and 0 at all others, enabling an explicit interpolating polynomial as a weighted sum.
- •An example illustrates unnormalized and normalized basis functions and the resulting interpolating polynomial that matches the sample points.