Physicist uses Stein Variational Gradient Descent to evenly light a Christmas tree
PTenigma · x · 2026-11-06
Paul Thompson turned his daughter's complaint about uneven Christmas tree lights into an SVGD demo: model the lights as particles on a cone surface and use kernel Stein variational updates to push them toward uniform density, under the constraint that lights stay on an equispaced 1-D string (spacing ≤ L cm). He analyzes the covering radius objective — hexagonal lattices are optimal in the flat plane but not on a cone, so the obvious serpentine layout isn't theoretical optimum. A playful application of a technique more commonly seen in generative modeling.
Related event: Mathematician uses SVGD to evenly distribute Christmas tree lights(2 posts)→
More from Fun
- Senior devs refusing last minute small changes: a meme every engineer knows — _jaydeepkarale · 2026-11-06
- RAM Is So Expensive People Are Joking About Building Digital RAM in Minecraft — djcows · 2026-11-06
- METR's no-acceleration verdict becomes a punchline against neocloud IPOs — teodorio · 2026-11-06
- Anthropic team lead hailed as a Marvel character: MD, JD, PhD and an art history MS — himanshustwts · 2026-11-06
- 'RIP Cybertruck' meme claims 2023-2027 run as Tesla pivots to Cybercab — mark_k · 2026-11-06
- GPT-6.1 Sol and GPT-6 Astra are all you need — until GPT-6.2 — Angaisb_ · 2026-11-06