Hacker News

...

jivaprime

10

Karma

2025-11-24

Created

About Me

math, algorithm, AI

Recent Activity

  • Submitted: "Show HN: Per-instance TSP Solver with No Pre-training (1.66% gap on d1291)"

    21 points5 comments

    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 achie...

  • Submitted: "CPU-only PPO solving TSPLIB lin318 in 20 mins (0.08% gap)"

    5 points0 comments

    Hi all

    I’ve put together a repo demonstrating how to train PPO directly on a single TSPLIB instance (lin318) from scratch—without pre-training or GPUs.

    Repo:https://github.com/jivaprime/TSP

    1. Experiment Setup

    Problem: TSPLIB lin318 (Opt: 42,029) & rd400

    Hardware: G...

HackerNews

About