nowfound

Life & fun · December 29, 2025

PI

Per-instance TSP Solver with No Pre-training (1.66% gap on d1291)

OP here. Most Deep Learning approaches for TSP rely on pre-training with large-scale datasets. I wanted to see if a solver could learn "on the fly" for a specific instance without any priors from other problems. I built a solver using PPO that learns from scratch per instance. It achieved a 1.66% gap on TSPLIB d1291 in about 5.6 hours on a single A100. The Core Idea: My hypothesis was that while optimal solutions are mostly composed of 'minimum edges' (nearest neighbors), the actual difficulty comes from a small number of 'exception edges' outside of that local scope. Instead of…

Per-instance TSP Solver with No Pre-training (1.66% gap on d1291) launched on December 29, 2025 with 21 votes, #573 of 11,628 launches that month and more than 90% of that year's launches.

Alternativestop 5% of December 2025

In plain words

This is a traveling salesman problem solver that learns independently for each problem instance using reinforcement learning, without relying on pre-trained models from other datasets. Built with PPO, it achieved a 1.66% optimality gap on a standard benchmark instance, solving it in about 5.6 hours on a single GPU. The approach is designed for researchers and practitioners working on combinatorial optimization who want to explore instance-specific learning strategies that leverage geometric structure rather than cross-problem patterns.

written from the facts on this page · September 2026

From the sources

In the maker’s words, at launch

OP here. Most Deep Learning approaches for TSP rely on pre-training with large-scale datasets. I wanted to see if a solver could learn "on the fly" for a specific instance without any priors from other problems. I built a solver using PPO that learns from scratch per instance. It achieved a 1.66% gap on TSPLIB d1291…

Does a similar job

all alternatives →
  • Watch a neural net learn to play SnakeMay 2026 · ppo.gradexp.xyz · ▲205

    In browser PPO training demo, made possible by tinygrad: TinyJit -> WebGPU kernels. Requires WebGPU.

  • How I topped the HuggingFace open LLM leaderboard on two gaming GPUsMar 2026 · dnhkng.github.io · ▲495

    I found that duplicating a specific block of 7 middle layers in Qwen2-72B, without modifying any weights, improved performance across all Open LLM Leaderboard benchmarks and took #1. As of 2026, the top 4 models on that leaderboard are still descendants. The weird finding: single-layer duplication does nothing. Too few layers, nothing. Too many, it gets worse. Only circuit-sized blocks of ~7 layers work. This suggests pretraining carves out discrete functional circuits in the layer stack that only work when preserved whole. The whole thing was developed on 2x RTX 4090s in my basement. I'm…

  • Interactive visual solver for the traveling salesman problem2019 · tspvis.com · ▲113

    Interactive solver for the traveling salesman problem to visualize different algorithms. Includes various Heuristic and Exhaustive algorithms.

  • BP
    Beating Pokemon Red with RL and <10M Parameters2025 · drubinstein.github.io · ▲183

    Hi everyone! After spending hundreds of hours, we're excited to finally share our progress in developing a reinforcement learning system to beat Pokémon Red. Our system successfully completes the game using a policy under 10M parameters, PPO, and a few novel techniques. With the release of Claude Plays Pokémon, now feels like the perfect time to showcase our work. We'd love to get feedback!

  • Fast Deep Reinforcement Learning Course2022 · courses.dibya.online · ▲148

    I worked on this applied Deep Reinforcement Learning course for the better part of 2021. I made a Datacamp course [0] before, and this served as my inspiration to make an applied Deep RL series. Normally, Deep RL courses teach a lot of mathematically involved theory. You get the practical applications near the end (if at all). I have tried to turn that on its head. In the top-down approach, you learn practical skills first, then go deeper later. This is much more fun. This course (the first in a planned multi-part series) shows how to use the Deep Reinforcement Learning framework RLlib to…

  • Boldly go where Gradient Descent has never gone before with DiscoGrad2024 · github.com · ▲232

    Trying to do gradient descent using automatic differentiation over branchy programs? Or to combine them with neural networks for end-to-end training? Then this might be interesting to you. We develped DiscoGrad, a tool for automatic differentiation through C++ programs involving input-dependent control flow (e.g., "if (f(x) < c) { ... }", differentiating wrt. x) and randomness. Our initial motivation was to enable the use of gradient descent with simulations, which often rely heavily on such discrete branching. The latter makes plain autodiff mostly useless, since it can only account for the…

More life & fun this month

the category →

Source: Show HN post ↗ · launch text quoted in part

Launched alongside, December 2025

the whole month →
  • GP
    Gemini Pro 3 imagines the HN front page 10 years from now▲3,346

    Life & fun · Dec 2025 · dosaygo-studio.github.io

  • Jmail – Google Suite for Epstein files▲1,557

    Hi everyone! My name's Luke and I made the original Jmail here alongside Riley Walz. We had a ton of friends collaborate on building out more of the app suite last night in lieue of DOJ's "Epstein files" release. Please AMA!

    Life & fun · Dec 2025 · jmail.world

  • Loki.Build▲743

    Design and ship studio-grade landing pages with AI

    AI · Dec 2025 · loki.build

  • PlanEat AI▲733

    AI turns your health goals into a 7-day menu & grocery list

    AI · Dec 2025 · planeatai.com

  • 22 GB of Hacker News in SQLite▲730

    Community, All the HN belong to you. This is an archive of hacker news that fits in your browser. When I made HN Made of Primes I realized I could probably do this offline sqlite&#x2F;wasm thing with the whole GBs of archive. The whole dataset. So I tried it, and this is it. Have Hacker News on your device. Go to this repo (https:&#x2F;&#x2F;github.com&#x2F;DOSAYGO-STUDIO&#x2F;HackerBook): you can download it. Big Query -> ETL -> npx serve docs - that's it. 20 years of HN arguments and beauty, can be yours forever. So they'll never die. Ever. It's the unkillable static archive of HN and it's…

    Dev tools · Dec 2025 · hackerbook.dosaygo.com

  • CyberCut AI▲723

    AI video studio for viral social clips

    AI · Dec 2025 · cybercut.ai